diff --git a/.gitignore b/.gitignore index c18dd8d..252b02d 100644 --- a/.gitignore +++ b/.gitignore @@ -1 +1,2 @@ __pycache__/ +suno_bark.egg-info/ diff --git a/LICENSE b/LICENSE index d1bbe80..11aac3a 100644 --- a/LICENSE +++ b/LICENSE @@ -1,400 +1,21 @@ - -Attribution-NonCommercial 4.0 International - -======================================================================= - -Creative Commons Corporation ("Creative Commons") is not a law firm and -does not provide legal services or legal advice. Distribution of -Creative Commons public licenses does not create a lawyer-client or -other relationship. Creative Commons makes its licenses and related -information available on an "as-is" basis. Creative Commons gives no -warranties regarding its licenses, any material licensed under their -terms and conditions, or any related information. Creative Commons -disclaims all liability for damages resulting from their use to the -fullest extent possible. - -Using Creative Commons Public Licenses - -Creative Commons public licenses provide a standard set of terms and -conditions that creators and other rights holders may use to share -original works of authorship and other material subject to copyright -and certain other rights specified in the public license below. The -following considerations are for informational purposes only, are not -exhaustive, and do not form part of our licenses. - - Considerations for licensors: Our public licenses are - intended for use by those authorized to give the public - permission to use material in ways otherwise restricted by - copyright and certain other rights. Our licenses are - irrevocable. Licensors should read and understand the terms - and conditions of the license they choose before applying it. - Licensors should also secure all rights necessary before - applying our licenses so that the public can reuse the - material as expected. Licensors should clearly mark any - material not subject to the license. This includes other CC- - licensed material, or material used under an exception or - limitation to copyright. More considerations for licensors: - wiki.creativecommons.org/Considerations_for_licensors - - Considerations for the public: By using one of our public - licenses, a licensor grants the public permission to use the - licensed material under specified terms and conditions. If - the licensor's permission is not necessary for any reason--for - example, because of any applicable exception or limitation to - copyright--then that use is not regulated by the license. Our - licenses grant only permissions under copyright and certain - other rights that a licensor has authority to grant. Use of - the licensed material may still be restricted for other - reasons, including because others have copyright or other - rights in the material. A licensor may make special requests, - such as asking that all changes be marked or described. - Although not required by our licenses, you are encouraged to - respect those requests where reasonable. More_considerations - for the public: - wiki.creativecommons.org/Considerations_for_licensees - -======================================================================= - -Creative Commons Attribution-NonCommercial 4.0 International Public -License - -By exercising the Licensed Rights (defined below), You accept and agree -to be bound by the terms and conditions of this Creative Commons -Attribution-NonCommercial 4.0 International Public License ("Public -License"). To the extent this Public License may be interpreted as a -contract, You are granted the Licensed Rights in consideration of Your -acceptance of these terms and conditions, and the Licensor grants You -such rights in consideration of benefits the Licensor receives from -making the Licensed Material available under these terms and -conditions. - -Section 1 -- Definitions. - - a. Adapted Material means material subject to Copyright and Similar - Rights that is derived from or based upon the Licensed Material - and in which the Licensed Material is translated, altered, - arranged, transformed, or otherwise modified in a manner requiring - permission under the Copyright and Similar Rights held by the - Licensor. For purposes of this Public License, where the Licensed - Material is a musical work, performance, or sound recording, - Adapted Material is always produced where the Licensed Material is - synched in timed relation with a moving image. - - b. Adapter's License means the license You apply to Your Copyright - and Similar Rights in Your contributions to Adapted Material in - accordance with the terms and conditions of this Public License. - - c. Copyright and Similar Rights means copyright and/or similar rights - closely related to copyright including, without limitation, - performance, broadcast, sound recording, and Sui Generis Database - Rights, without regard to how the rights are labeled or - categorized. For purposes of this Public License, the rights - specified in Section 2(b)(1)-(2) are not Copyright and Similar - Rights. - d. Effective Technological Measures means those measures that, in the - absence of proper authority, may not be circumvented under laws - fulfilling obligations under Article 11 of the WIPO Copyright - Treaty adopted on December 20, 1996, and/or similar international - agreements. - - e. Exceptions and Limitations means fair use, fair dealing, and/or - any other exception or limitation to Copyright and Similar Rights - that applies to Your use of the Licensed Material. - - f. Licensed Material means the artistic or literary work, database, - or other material to which the Licensor applied this Public - License. - - g. Licensed Rights means the rights granted to You subject to the - terms and conditions of this Public License, which are limited to - all Copyright and Similar Rights that apply to Your use of the - Licensed Material and that the Licensor has authority to license. - - h. Licensor means the individual(s) or entity(ies) granting rights - under this Public License. - - i. NonCommercial means not primarily intended for or directed towards - commercial advantage or monetary compensation. For purposes of - this Public License, the exchange of the Licensed Material for - other material subject to Copyright and Similar Rights by digital - file-sharing or similar means is NonCommercial provided there is - no payment of monetary compensation in connection with the - exchange. - - j. Share means to provide material to the public by any means or - process that requires permission under the Licensed Rights, such - as reproduction, public display, public performance, distribution, - dissemination, communication, or importation, and to make material - available to the public including in ways that members of the - public may access the material from a place and at a time - individually chosen by them. - - k. Sui Generis Database Rights means rights other than copyright - resulting from Directive 96/9/EC of the European Parliament and of - the Council of 11 March 1996 on the legal protection of databases, - as amended and/or succeeded, as well as other essentially - equivalent rights anywhere in the world. - - l. You means the individual or entity exercising the Licensed Rights - under this Public License. Your has a corresponding meaning. - -Section 2 -- Scope. - - a. License grant. - - 1. Subject to the terms and conditions of this Public License, - the Licensor hereby grants You a worldwide, royalty-free, - non-sublicensable, non-exclusive, irrevocable license to - exercise the Licensed Rights in the Licensed Material to: - - a. reproduce and Share the Licensed Material, in whole or - in part, for NonCommercial purposes only; and - - b. produce, reproduce, and Share Adapted Material for - NonCommercial purposes only. - - 2. Exceptions and Limitations. For the avoidance of doubt, where - Exceptions and Limitations apply to Your use, this Public - License does not apply, and You do not need to comply with - its terms and conditions. - - 3. Term. The term of this Public License is specified in Section - 6(a). - - 4. Media and formats; technical modifications allowed. The - Licensor authorizes You to exercise the Licensed Rights in - all media and formats whether now known or hereafter created, - and to make technical modifications necessary to do so. The - Licensor waives and/or agrees not to assert any right or - authority to forbid You from making technical modifications - necessary to exercise the Licensed Rights, including - technical modifications necessary to circumvent Effective - Technological Measures. For purposes of this Public License, - simply making modifications authorized by this Section 2(a) - (4) never produces Adapted Material. - - 5. Downstream recipients. - - a. Offer from the Licensor -- Licensed Material. Every - recipient of the Licensed Material automatically - receives an offer from the Licensor to exercise the - Licensed Rights under the terms and conditions of this - Public License. - - b. No downstream restrictions. You may not offer or impose - any additional or different terms or conditions on, or - apply any Effective Technological Measures to, the - Licensed Material if doing so restricts exercise of the - Licensed Rights by any recipient of the Licensed - Material. - - 6. No endorsement. Nothing in this Public License constitutes or - may be construed as permission to assert or imply that You - are, or that Your use of the Licensed Material is, connected - with, or sponsored, endorsed, or granted official status by, - the Licensor or others designated to receive attribution as - provided in Section 3(a)(1)(A)(i). - - b. Other rights. - - 1. Moral rights, such as the right of integrity, are not - licensed under this Public License, nor are publicity, - privacy, and/or other similar personality rights; however, to - the extent possible, the Licensor waives and/or agrees not to - assert any such rights held by the Licensor to the limited - extent necessary to allow You to exercise the Licensed - Rights, but not otherwise. - - 2. Patent and trademark rights are not licensed under this - Public License. - - 3. To the extent possible, the Licensor waives any right to - collect royalties from You for the exercise of the Licensed - Rights, whether directly or through a collecting society - under any voluntary or waivable statutory or compulsory - licensing scheme. In all other cases the Licensor expressly - reserves any right to collect such royalties, including when - the Licensed Material is used other than for NonCommercial - purposes. - -Section 3 -- License Conditions. - -Your exercise of the Licensed Rights is expressly made subject to the -following conditions. - - a. Attribution. - - 1. If You Share the Licensed Material (including in modified - form), You must: - - a. retain the following if it is supplied by the Licensor - with the Licensed Material: - - i. identification of the creator(s) of the Licensed - Material and any others designated to receive - attribution, in any reasonable manner requested by - the Licensor (including by pseudonym if - designated); - - ii. a copyright notice; - - iii. a notice that refers to this Public License; - - iv. a notice that refers to the disclaimer of - warranties; - - v. a URI or hyperlink to the Licensed Material to the - extent reasonably practicable; - - b. indicate if You modified the Licensed Material and - retain an indication of any previous modifications; and - - c. indicate the Licensed Material is licensed under this - Public License, and include the text of, or the URI or - hyperlink to, this Public License. - - 2. You may satisfy the conditions in Section 3(a)(1) in any - reasonable manner based on the medium, means, and context in - which You Share the Licensed Material. For example, it may be - reasonable to satisfy the conditions by providing a URI or - hyperlink to a resource that includes the required - information. - - 3. If requested by the Licensor, You must remove any of the - information required by Section 3(a)(1)(A) to the extent - reasonably practicable. - - 4. If You Share Adapted Material You produce, the Adapter's - License You apply must not prevent recipients of the Adapted - Material from complying with this Public License. - -Section 4 -- Sui Generis Database Rights. - -Where the Licensed Rights include Sui Generis Database Rights that -apply to Your use of the Licensed Material: - - a. for the avoidance of doubt, Section 2(a)(1) grants You the right - to extract, reuse, reproduce, and Share all or a substantial - portion of the contents of the database for NonCommercial purposes - only; - - b. if You include all or a substantial portion of the database - contents in a database in which You have Sui Generis Database - Rights, then the database in which You have Sui Generis Database - Rights (but not its individual contents) is Adapted Material; and - - c. You must comply with the conditions in Section 3(a) if You Share - all or a substantial portion of the contents of the database. - -For the avoidance of doubt, this Section 4 supplements and does not -replace Your obligations under this Public License where the Licensed -Rights include other Copyright and Similar Rights. - -Section 5 -- Disclaimer of Warranties and Limitation of Liability. - - a. UNLESS OTHERWISE SEPARATELY UNDERTAKEN BY THE LICENSOR, TO THE - EXTENT POSSIBLE, THE LICENSOR OFFERS THE LICENSED MATERIAL AS-IS - AND AS-AVAILABLE, AND MAKES NO REPRESENTATIONS OR WARRANTIES OF - ANY KIND CONCERNING THE LICENSED MATERIAL, WHETHER EXPRESS, - IMPLIED, STATUTORY, OR OTHER. THIS INCLUDES, WITHOUT LIMITATION, - WARRANTIES OF TITLE, MERCHANTABILITY, FITNESS FOR A PARTICULAR - PURPOSE, NON-INFRINGEMENT, ABSENCE OF LATENT OR OTHER DEFECTS, - ACCURACY, OR THE PRESENCE OR ABSENCE OF ERRORS, WHETHER OR NOT - KNOWN OR DISCOVERABLE. WHERE DISCLAIMERS OF WARRANTIES ARE NOT - ALLOWED IN FULL OR IN PART, THIS DISCLAIMER MAY NOT APPLY TO YOU. - - b. TO THE EXTENT POSSIBLE, IN NO EVENT WILL THE LICENSOR BE LIABLE - TO YOU ON ANY LEGAL THEORY (INCLUDING, WITHOUT LIMITATION, - NEGLIGENCE) OR OTHERWISE FOR ANY DIRECT, SPECIAL, INDIRECT, - INCIDENTAL, CONSEQUENTIAL, PUNITIVE, EXEMPLARY, OR OTHER LOSSES, - COSTS, EXPENSES, OR DAMAGES ARISING OUT OF THIS PUBLIC LICENSE OR - USE OF THE LICENSED MATERIAL, EVEN IF THE LICENSOR HAS BEEN - ADVISED OF THE POSSIBILITY OF SUCH LOSSES, COSTS, EXPENSES, OR - DAMAGES. WHERE A LIMITATION OF LIABILITY IS NOT ALLOWED IN FULL OR - IN PART, THIS LIMITATION MAY NOT APPLY TO YOU. - - c. The disclaimer of warranties and limitation of liability provided - above shall be interpreted in a manner that, to the extent - possible, most closely approximates an absolute disclaimer and - waiver of all liability. - -Section 6 -- Term and Termination. - - a. This Public License applies for the term of the Copyright and - Similar Rights licensed here. However, if You fail to comply with - this Public License, then Your rights under this Public License - terminate automatically. - - b. Where Your right to use the Licensed Material has terminated under - Section 6(a), it reinstates: - - 1. automatically as of the date the violation is cured, provided - it is cured within 30 days of Your discovery of the - violation; or - - 2. upon express reinstatement by the Licensor. - - For the avoidance of doubt, this Section 6(b) does not affect any - right the Licensor may have to seek remedies for Your violations - of this Public License. - - c. For the avoidance of doubt, the Licensor may also offer the - Licensed Material under separate terms or conditions or stop - distributing the Licensed Material at any time; however, doing so - will not terminate this Public License. - - d. Sections 1, 5, 6, 7, and 8 survive termination of this Public - License. - -Section 7 -- Other Terms and Conditions. - - a. The Licensor shall not be bound by any additional or different - terms or conditions communicated by You unless expressly agreed. - - b. Any arrangements, understandings, or agreements regarding the - Licensed Material not stated herein are separate from and - independent of the terms and conditions of this Public License. - -Section 8 -- Interpretation. - - a. For the avoidance of doubt, this Public License does not, and - shall not be interpreted to, reduce, limit, restrict, or impose - conditions on any use of the Licensed Material that could lawfully - be made without permission under this Public License. - - b. To the extent possible, if any provision of this Public License is - deemed unenforceable, it shall be automatically reformed to the - minimum extent necessary to make it enforceable. If the provision - cannot be reformed, it shall be severed from this Public License - without affecting the enforceability of the remaining terms and - conditions. - - c. No term or condition of this Public License will be waived and no - failure to comply consented to unless expressly agreed to by the - Licensor. - - d. Nothing in this Public License constitutes or may be interpreted - as a limitation upon, or waiver of, any privileges and immunities - that apply to the Licensor or You, including from the legal - processes of any jurisdiction or authority. - -======================================================================= - -Creative Commons is not a party to its public -licenses. Notwithstanding, Creative Commons may elect to apply one of -its public licenses to material it publishes and in those instances -will be considered the โ€œLicensor.โ€ The text of the Creative Commons -public licenses is dedicated to the public domain under the CC0 Public -Domain Dedication. Except for the limited purpose of indicating that -material is shared under a Creative Commons public license or as -otherwise permitted by the Creative Commons policies published at -creativecommons.org/policies, Creative Commons does not authorize the -use of the trademark "Creative Commons" or any other trademark or logo -of Creative Commons without its prior written consent including, -without limitation, in connection with any unauthorized modifications -to any of its public licenses or any other arrangements, -understandings, or agreements concerning use of licensed material. For -the avoidance of doubt, this paragraph does not form part of the -public licenses. - -Creative Commons may be contacted at creativecommons.org. +MIT License + +Copyright (c) Suno, Inc + +Permission is hereby granted, free of charge, to any person obtaining a copy +of this software and associated documentation files (the "Software"), to deal +in the Software without restriction, including without limitation the rights +to use, copy, modify, merge, publish, distribute, sublicense, and/or sell +copies of the Software, and to permit persons to whom the Software is +furnished to do so, subject to the following conditions: + +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. diff --git a/README.md b/README.md index a44887b..4b35eaf 100644 --- a/README.md +++ b/README.md @@ -1,27 +1,53 @@ # ๐Ÿถ Bark - +[![](https://dcbadge.vercel.app/api/server/J2B2vsjKuE?style=flat&compact=True)](https://discord.gg/J2B2vsjKuE) [![Twitter](https://img.shields.io/twitter/url/https/twitter.com/OnusFM.svg?style=social&label=@OnusFM)](https://twitter.com/OnusFM) -[![](https://dcbadge.vercel.app/api/server/J2B2vsjKuE?compact=true&style=flat&)](https://discord.gg/J2B2vsjKuE) + +[Examples](https://suno-ai.notion.site/Bark-Examples-5edae8b02a604b54a42244ba45ebc2e2) โ€ข [Suno Studio Waitlist](https://3os84zs17th.typeform.com/suno-studio) โ€ข [Updates](#-updates) โ€ข [How to Use](#-usage-in-python) โ€ข [Installation](#-installation) โ€ข [FAQ](#-faq) -[Examples](https://suno-ai.notion.site/Bark-Examples-5edae8b02a604b54a42244ba45ebc2e2) | [Model Card](./model-card.md) | [Playground Waitlist](https://3os84zs17th.typeform.com/suno-studio) - -Bark is a transformer-based text-to-audio model created by [Suno](https://suno.ai). Bark can generate highly realistic, multilingual speech as well as other audio - including music, background noise and simple sound effects. The model can also produce nonverbal communications like laughing, sighing and crying. To support the research community, we are providing access to pretrained model checkpoints ready for inference. - +[//]:
(vertical spaces around image) +

- +

+
-## ๐Ÿ”Š Demos +Bark is a transformer-based text-to-audio model created by [Suno](https://suno.ai). Bark can generate highly realistic, multilingual speech as well as other audio - including music, background noise and simple sound effects. The model can also produce nonverbal communications like laughing, sighing and crying. To support the research community, we are providing access to pretrained model checkpoints, which are ready for inference and available for commercial use. -[![Open in Spaces](https://img.shields.io/badge/๐Ÿค—-Open%20In%20Spaces-blue.svg)](https://huggingface.co/spaces/suno/bark) +## โš  Disclaimer +Bark was developed for research purposes. It is not a conventional text-to-speech model but instead a fully generative text-to-audio model, which can deviate in unexpected ways from provided prompts. Suno does not take responsibility for any output generated. Use at your own risk, and please act responsibly. + +## ๐ŸŽง Demos + +[![Open in Spaces](https://img.shields.io/badge/๐Ÿค—-Open%20in%20Spaces-blue.svg)](https://huggingface.co/spaces/suno/bark) +[![Open on Replicate](https://img.shields.io/badge/ยฎ๏ธ-Open%20on%20Replicate-blue.svg)](https://replicate.com/suno-ai/bark) [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1eJfA2XUa-mXwdMy7DoYKVYHI1iTd9Vkt?usp=sharing) -## ๐Ÿค– Usage +## ๐Ÿš€ Updates + +**2023.05.01** +- ยฉ๏ธ Bark is now licensed under the MIT License, meaning it's now available for commercial use! +- โšก 2x speed-up on GPU. 10x speed-up on CPU. We also added an option for a smaller version of Bark, which offers additional speed-up with the trade-off of slightly lower quality. +- ๐Ÿ“• [Long-form generation](notebooks/long_form_generation.ipynb), voice consistency enhancements and other examples are now documented in a new [notebooks](./notebooks) section. +- ๐Ÿ‘ฅ We created a [voice prompt library](https://suno-ai.notion.site/8b8e8749ed514b0cbf3f699013548683?v=bc67cff786b04b50b3ceb756fd05f68c). We hope this resource helps you find useful prompts for your use cases! You can also join us on [Discord](https://discord.gg/J2B2vsjKuE), where the community actively shares useful prompts in the **#audio-prompts** channel. +- ๐Ÿ’ฌ Growing community support and access to new features here: + + [![](https://dcbadge.vercel.app/api/server/J2B2vsjKuE)](https://discord.gg/J2B2vsjKuE) + +- ๐Ÿ’พ You can now use Bark with GPUs that have low VRAM (<4GB). + +**2023.04.20** +- ๐Ÿถ Bark release! + +## ๐Ÿ Usage in Python + +
+

๐Ÿช‘ Basics

 ```python from bark import SAMPLE_RATE, generate_audio, preload_models +from scipy.io.wavfile import write as write_wav from IPython.display import Audio # download and load all models @@ -34,38 +60,42 @@ text_prompt = """ """ audio_array = generate_audio(text_prompt) +# save audio to disk +write_wav("bark_generation.wav", SAMPLE_RATE, audio_array) + # play text in notebook Audio(audio_array, rate=SAMPLE_RATE) ``` [pizza.webm](https://user-images.githubusercontent.com/5068315/230490503-417e688d-5115-4eee-9550-b46a2b465ee3.webm) +
-To save `audio_array` as a WAV file: - -```python -from scipy.io.wavfile import write as write_wav - -write_wav("/path/to/audio.wav", SAMPLE_RATE, audio_array) -``` - -### ๐ŸŒŽ Foreign Language - +
+

๐ŸŒŽ Foreign Language

 +
Bark supports various languages out-of-the-box and automatically determines language from input text. When prompted with code-switched text, Bark will attempt to employ the native accent for the respective languages. English quality is best for the time being, and we expect other languages to further improve with scaling. +
+
```python + text_prompt = """ - Buenos dรญas Miguel. Tu colega piensa que tu alemรกn es extremadamente malo. - But I suppose your english isn't terrible. + ์ถ”์„์€ ๋‚ด๊ฐ€ ๊ฐ€์žฅ ์ข‹์•„ํ•˜๋Š” ๋ช…์ ˆ์ด๋‹ค. ๋‚˜๋Š” ๋ฉฐ์น  ๋™์•ˆ ํœด์‹์„ ์ทจํ•˜๊ณ  ์นœ๊ตฌ ๋ฐ ๊ฐ€์กฑ๊ณผ ์‹œ๊ฐ„์„ ๋ณด๋‚ผ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. """ audio_array = generate_audio(text_prompt) ``` +[suno_korean.webm](https://user-images.githubusercontent.com/32879321/235313033-dc4477b9-2da0-4b94-9c8b-a8c2d8f5bb5e.webm) + +*Note: since Bark recognizes languages automatically from input text, it is possible to use for example a german history prompt with english text. This usually leads to english audio with a german accent.* -[miguel.webm](https://user-images.githubusercontent.com/5068315/230684752-10baadfe-1e7c-46a2-8323-43282aef2c8c.webm) - -### ๐ŸŽถ Music +
+
+

๐ŸŽถ Music

 Bark can generate all types of audio, and, in principle, doesn't see a difference between speech and music. Sometimes Bark chooses to generate text as music, but you can help it out by adding music notes around your lyrics. +
+
```python text_prompt = """ @@ -73,39 +103,45 @@ text_prompt = """ """ audio_array = generate_audio(text_prompt) ``` - [lion.webm](https://user-images.githubusercontent.com/5068315/230684766-97f5ea23-ad99-473c-924b-66b6fab24289.webm) +
-### ๐ŸŽค Voice Presets and Voice/Audio Cloning +
+

๐ŸŽค Voice Presets

 + +Bark supports 100+ speaker presets across [supported languages](#supported-languages). You can browse the library of speaker presets [here](https://suno-ai.notion.site/8b8e8749ed514b0cbf3f699013548683?v=bc67cff786b04b50b3ceb756fd05f68c), or in the [code](bark/assets/prompts). The community also often shares presets in [Discord](https://discord.gg/J2B2vsjKuE). -Bark has the capability to fully clone voices - including tone, pitch, emotion and prosody. The model also attempts to preserve music, ambient noise, etc. from input audio. However, to mitigate misuse of this technology, we limit the audio history prompts to a limited set of Suno-provided, fully synthetic options to choose from for each language. Specify following the pattern: `{lang_code}_speaker_{0-9}`. +Bark tries to match the tone, pitch, emotion and prosody of a given preset, but does not currently support custom voice cloning. The model also attempts to preserve music, ambient noise, etc. +
+
```python text_prompt = """ I have a silky smooth voice, and today I will tell you about the exercise regimen of the common sloth. """ -audio_array = generate_audio(text_prompt, history_prompt="en_speaker_1") +audio_array = generate_audio(text_prompt, history_prompt="v2/en_speaker_1") ``` - [sloth.webm](https://user-images.githubusercontent.com/5068315/230684883-a344c619-a560-4ff5-8b99-b4463a34487b.webm) +
-*Note: since Bark recognizes languages automatically from input text, it is possible to use for example a german history prompt with english text. This usually leads to english audio with a german accent.* +### Generating Longer Audio + +By default, `generate_audio` works well with around 13 seconds of spoken text. For an example of how to do long-form generation, see this [example notebook](notebooks/long_form_generation.ipynb). -### ๐Ÿ‘ฅ Speaker Prompts +
+Click to toggle example long-form generations (from the example notebook) -You can provide certain speaker prompts such as NARRATOR, MAN, WOMAN, etc. Please note that these are not always respected, especially if a conflicting audio history prompt is given. +[dialog.webm](https://user-images.githubusercontent.com/2565833/235463539-f57608da-e4cb-4062-8771-148e29512b01.webm) + +[longform_advanced.webm](https://user-images.githubusercontent.com/2565833/235463547-1c0d8744-269b-43fe-9630-897ea5731652.webm) + +[longform_basic.webm](https://user-images.githubusercontent.com/2565833/235463559-87efe9f8-a2db-4d59-b764-57db83f95270.webm) + +
-```python -text_prompt = """ - WOMAN: I would like an oatmilk latte please. - MAN: Wow, that's expensive! -""" -audio_array = generate_audio(text_prompt) -``` -[latte.webm](https://user-images.githubusercontent.com/5068315/230684864-12d101a1-a726-471d-9d56-d18b108efcb8.webm) ## ๐Ÿ’ป Installation @@ -120,20 +156,23 @@ or git clone https://github.com/suno-ai/bark cd bark && pip install . ``` +*Note: Do NOT use 'pip install bark'. It installs a different package, which is not managed by Suno.* + ## ๐Ÿ› ๏ธ Hardware and Inference Speed Bark has been tested and works on both CPU and GPU (`pytorch 2.0+`, CUDA 11.7 and CUDA 12.0). -Running Bark requires running >100M parameter transformer models. -On modern GPUs and PyTorch nightly, Bark can generate audio in roughly realtime. On older GPUs, default colab, or CPU, inference time might be 10-100x slower. -If you don't have new hardware available or if you want to play with bigger versions of our models, you can also sign up for early access to our model playground [here](https://3os84zs17th.typeform.com/suno-studio). +On enterprise GPUs and PyTorch nightly, Bark can generate audio in roughly real-time. On older GPUs, default colab, or CPU, inference time might be significantly slower. For older GPUs or CPU you might want to consider using smaller models. Details can be found in out tutorial sections here. + +The full version of Bark requires around 12GB of VRAM to hold everything on GPU at the same time. +To use a smaller version of the models, which should fit into 8GB VRAM, set the environment flag `SUNO_USE_SMALL_MODELS=True`. + +If you don't have hardware available or if you want to play with bigger versions of our models, you can also sign up for early access to our model playground [here](https://3os84zs17th.typeform.com/suno-studio). ## โš™๏ธ Details -Similar to [Vall-E](https://arxiv.org/abs/2301.02111) and some other amazing work in the field, Bark uses GPT-style -models to generate audio from scratch. Different from Vall-E, the initial text prompt is embedded into high-level semantic tokens without the use of phonemes. It can therefore generalize to arbitrary instructions beyond speech that occur in the training data, such as music lyrics, sound effects or other non-speech sounds. A subsequent second model is used to convert the generated semantic tokens into audio codec tokens to generate the full waveform. To enable the community to use Bark via public code we used the fantastic -[EnCodec codec](https://github.com/facebookresearch/encodec) from Facebook to act as an audio representation. +Bark is fully generative tex-to-audio model devolved for research and demo purposes. It follows a GPT style architecture similar to [AudioLM](https://arxiv.org/abs/2209.03143) and [Vall-E](https://arxiv.org/abs/2301.02111) and a quantized Audio representation from [EnCodec](https://github.com/facebookresearch/encodec). It is not a conventional TTS model, but instead a fully generative text-to-audio model capable of deviating in unexpected ways from any given script. Different to previous approaches, the input text prompt is converted directly to audio without the intermediate use of phonemes. It can therefore generalize to arbitrary instructions beyond speech such as music lyrics, sound effects or other non-speech sounds. Below is a list of some known non-speech sounds, but we are finding more every day. Please let us know if you find patterns that work particularly well on [Discord](https://discord.gg/J2B2vsjKuE)! @@ -145,10 +184,10 @@ Below is a list of some known non-speech sounds, but we are finding more every d - `[clears throat]` - `โ€”` or `...` for hesitations - `โ™ช` for song lyrics -- capitalization for emphasis of a word -- `MAN/WOMAN:` for bias towards speaker +- CAPITALIZATION for emphasis of a word +- `[MAN]` and `[WOMAN]` to bias Bark toward male and female speakers, respectively -**Supported Languages** +### Supported Languages | Language | Status | | --- | --- | @@ -165,22 +204,21 @@ Below is a list of some known non-speech sounds, but we are finding more every d | Russian (ru) | โœ… | | Turkish (tr) | โœ… | | Chinese, simplified (zh) | โœ… | -| Arabic | Coming soon! | -| Bengali | Coming soon! | -| Telugu | Coming soon! | + +Requests for future language support [here](https://github.com/suno-ai/bark/discussions/111) or in the **#forums** channel on [Discord](https://discord.com/invite/J2B2vsjKuE). ## ๐Ÿ™ Appreciation - [nanoGPT](https://github.com/karpathy/nanoGPT) for a dead-simple and blazing fast implementation of GPT-style models - [EnCodec](https://github.com/facebookresearch/encodec) for a state-of-the-art implementation of a fantastic audio codec -- [AudioLM](https://github.com/lucidrains/audiolm-pytorch) for very related training and inference code +- [AudioLM](https://github.com/lucidrains/audiolm-pytorch) for related training and inference code - [Vall-E](https://arxiv.org/abs/2301.02111), [AudioLM](https://arxiv.org/abs/2209.03143) and many other ground-breaking papers that enabled the development of Bark ## ยฉ License -Bark is licensed under a non-commercial license: CC-BY 4.0 NC. The Suno models themselves may be used commercially. However, this version of Bark uses `EnCodec` as a neural codec backend, which is licensed under a [non-commercial license](https://github.com/facebookresearch/encodec/blob/main/LICENSE). +Bark is licensed under the MIT License. -Please contact us at `bark@suno.ai` if you need access to a larger version of the model and/or a version of the model you can use commercially. +Please contact us at `bark@suno.ai` to request access to a larger version of the model. ## ๐Ÿ“ฑย Community @@ -193,12 +231,29 @@ Weโ€™re developing a playground for our models, including Bark. If you are interested, you can sign up for early access [here](https://3os84zs17th.typeform.com/suno-studio). -## FAQ +## โ“ FAQ #### How do I specify where models are downloaded and cached? +* Bark uses Hugging Face to download and store models. You can see find more info [here](https://huggingface.co/docs/huggingface_hub/package_reference/environment_variables#hfhome). -Use the `XDG_CACHE_HOME` env variable to override where models are downloaded and cached (otherwise defaults to a subdirectory of `~/.cache`). #### Bark's generations sometimes differ from my prompts. What's happening? +* Bark is a GPT-style model. As such, it may take some creative liberties in its generations, resulting in higher-variance model outputs than traditional text-to-speech approaches. -Bark is a GPT-style model. As such, it may take some creative liberties in its generations, resulting in higher-variance model outputs than traditional text-to-speech approaches. +#### What voices are supported by Bark? +* Bark supports 100+ speaker presets across [supported languages](#supported-languages). You can browse the library of speaker presets [here](https://suno-ai.notion.site/8b8e8749ed514b0cbf3f699013548683?v=bc67cff786b04b50b3ceb756fd05f68c). The community also shares presets in [Discord](https://discord.gg/J2B2vsjKuE). Bark also supports generating unique random voices that fit the input text. Bark does not currently support custom voice cloning. + +#### Why is the output limited to ~13-14 seconds? +* Bark is a GPT-style model, and its architecture/context window is optimized to output generations with roughly this length. + +#### How much VRAM do I need? +* The full version of Bark requires around 12Gb of memory to hold everything on GPU at the same time. However, even smaller cards down to ~2Gb work with some additional settings. Simply add the following code snippet before your generation: + +```python +import os +os.environ["SUNO_OFFLOAD_CPU"] = True +os.environ["SUNO_USE_SMALL_MODELS"] = True +``` + +#### My generated audio sounds like a 1980s phone call. What's happening? +* Bark generates audio from scratch. It is not meant to create only high-fidelity, studio-quality speech. Rather, outputs could be anything from perfect speech to multiple people arguing at a baseball game recorded with bad microphones. diff --git a/bark/api.py b/bark/api.py index e1c7556..7b646b7 100644 --- a/bark/api.py +++ b/bark/api.py @@ -1,4 +1,4 @@ -from typing import Optional +from typing import Dict, Optional, Union import numpy as np @@ -7,7 +7,7 @@ from .generation import codec_decode, generate_coarse, generate_fine, generate_t def text_to_semantic( text: str, - history_prompt: Optional[str] = None, + history_prompt: Optional[Union[Dict, str]] = None, temp: float = 0.7, silent: bool = False, ): @@ -34,7 +34,7 @@ def text_to_semantic( def semantic_to_waveform( semantic_tokens: np.ndarray, - history_prompt: Optional[str] = None, + history_prompt: Optional[Union[Dict, str]] = None, temp: float = 0.7, silent: bool = False, output_full: bool = False, @@ -85,7 +85,7 @@ def save_as_prompt(filepath, full_generation): def generate_audio( text: str, - history_prompt: Optional[str] = None, + history_prompt: Optional[Union[Dict, str]] = None, text_temp: float = 0.7, waveform_temp: float = 0.7, silent: bool = False, diff --git a/bark/assets/prompts/readme.md b/bark/assets/prompts/readme.md index 2844c37..b01ae91 100644 --- a/bark/assets/prompts/readme.md +++ b/bark/assets/prompts/readme.md @@ -1,5 +1,15 @@ # Example Prompts Data +## Version Two +The `v2` prompts are better engineered to follow text with a consistent voice. +To use them, simply include `v2` in the prompt. For example +```python +from bark import generate_audio +text_prompt = "madam I'm adam" +audio_array = generate_audio(text_prompt, history_prompt="v2/en_speaker_1") +``` + +## Prompt Format The provided data is in the .npz format, which is a file format used in Python for storing arrays and data. The data contains three arrays: semantic_prompt, coarse_prompt, and fine_prompt. ```semantic_prompt``` diff --git a/bark/assets/prompts/v2/de_speaker_0.npz b/bark/assets/prompts/v2/de_speaker_0.npz new file mode 100644 index 0000000..0a10369 Binary files /dev/null and b/bark/assets/prompts/v2/de_speaker_0.npz differ diff --git a/bark/assets/prompts/v2/de_speaker_1.npz b/bark/assets/prompts/v2/de_speaker_1.npz new file mode 100644 index 0000000..da143ea Binary files /dev/null and b/bark/assets/prompts/v2/de_speaker_1.npz differ diff --git a/bark/assets/prompts/v2/de_speaker_2.npz b/bark/assets/prompts/v2/de_speaker_2.npz new file mode 100644 index 0000000..26aef4d Binary files /dev/null and b/bark/assets/prompts/v2/de_speaker_2.npz differ diff --git a/bark/assets/prompts/v2/de_speaker_3.npz b/bark/assets/prompts/v2/de_speaker_3.npz new file mode 100644 index 0000000..3bb685e Binary files /dev/null and b/bark/assets/prompts/v2/de_speaker_3.npz differ diff --git a/bark/assets/prompts/v2/de_speaker_4.npz b/bark/assets/prompts/v2/de_speaker_4.npz new file mode 100644 index 0000000..7fae90c Binary files /dev/null and b/bark/assets/prompts/v2/de_speaker_4.npz differ diff --git a/bark/assets/prompts/v2/de_speaker_5.npz b/bark/assets/prompts/v2/de_speaker_5.npz new file mode 100644 index 0000000..c5f2527 Binary files /dev/null and b/bark/assets/prompts/v2/de_speaker_5.npz differ diff --git a/bark/assets/prompts/v2/de_speaker_6.npz b/bark/assets/prompts/v2/de_speaker_6.npz new file mode 100644 index 0000000..5b26055 Binary files /dev/null and b/bark/assets/prompts/v2/de_speaker_6.npz differ diff --git a/bark/assets/prompts/v2/de_speaker_7.npz b/bark/assets/prompts/v2/de_speaker_7.npz new file mode 100644 index 0000000..f320937 Binary files /dev/null and b/bark/assets/prompts/v2/de_speaker_7.npz differ diff --git a/bark/assets/prompts/v2/de_speaker_8.npz b/bark/assets/prompts/v2/de_speaker_8.npz new file mode 100644 index 0000000..09bfc68 Binary files /dev/null and b/bark/assets/prompts/v2/de_speaker_8.npz differ diff --git a/bark/assets/prompts/v2/de_speaker_9.npz b/bark/assets/prompts/v2/de_speaker_9.npz new file mode 100644 index 0000000..bbbef85 Binary files /dev/null and b/bark/assets/prompts/v2/de_speaker_9.npz differ diff --git a/bark/assets/prompts/v2/en_speaker_0.npz b/bark/assets/prompts/v2/en_speaker_0.npz new file mode 100644 index 0000000..1ee9aee Binary files /dev/null and b/bark/assets/prompts/v2/en_speaker_0.npz differ diff --git a/bark/assets/prompts/v2/en_speaker_1.npz b/bark/assets/prompts/v2/en_speaker_1.npz new file mode 100644 index 0000000..381f175 Binary files /dev/null and b/bark/assets/prompts/v2/en_speaker_1.npz differ diff --git a/bark/assets/prompts/v2/en_speaker_2.npz b/bark/assets/prompts/v2/en_speaker_2.npz new file mode 100644 index 0000000..015b2e5 Binary files /dev/null and b/bark/assets/prompts/v2/en_speaker_2.npz differ diff --git a/bark/assets/prompts/v2/en_speaker_3.npz b/bark/assets/prompts/v2/en_speaker_3.npz new file mode 100644 index 0000000..e77005a Binary files /dev/null and b/bark/assets/prompts/v2/en_speaker_3.npz differ diff --git a/bark/assets/prompts/v2/en_speaker_4.npz b/bark/assets/prompts/v2/en_speaker_4.npz new file mode 100644 index 0000000..c75e9e6 Binary files /dev/null and b/bark/assets/prompts/v2/en_speaker_4.npz differ diff --git a/bark/assets/prompts/v2/en_speaker_5.npz b/bark/assets/prompts/v2/en_speaker_5.npz new file mode 100644 index 0000000..6a82e9a Binary files /dev/null and b/bark/assets/prompts/v2/en_speaker_5.npz differ diff --git a/bark/assets/prompts/v2/en_speaker_6.npz b/bark/assets/prompts/v2/en_speaker_6.npz new file mode 100644 index 0000000..fd1d9a0 Binary files /dev/null and b/bark/assets/prompts/v2/en_speaker_6.npz differ diff --git a/bark/assets/prompts/v2/en_speaker_7.npz b/bark/assets/prompts/v2/en_speaker_7.npz new file mode 100644 index 0000000..3e91009 Binary files /dev/null and b/bark/assets/prompts/v2/en_speaker_7.npz differ diff --git a/bark/assets/prompts/v2/en_speaker_8.npz b/bark/assets/prompts/v2/en_speaker_8.npz new file mode 100644 index 0000000..5e5077b Binary files /dev/null and b/bark/assets/prompts/v2/en_speaker_8.npz differ diff --git a/bark/assets/prompts/v2/en_speaker_9.npz b/bark/assets/prompts/v2/en_speaker_9.npz new file mode 100644 index 0000000..c8798ed Binary files /dev/null and b/bark/assets/prompts/v2/en_speaker_9.npz differ diff --git a/bark/assets/prompts/v2/es_speaker_0.npz b/bark/assets/prompts/v2/es_speaker_0.npz new file mode 100644 index 0000000..11877a1 Binary files /dev/null and b/bark/assets/prompts/v2/es_speaker_0.npz differ diff --git a/bark/assets/prompts/v2/es_speaker_1.npz b/bark/assets/prompts/v2/es_speaker_1.npz new file mode 100644 index 0000000..1daabec Binary files /dev/null and b/bark/assets/prompts/v2/es_speaker_1.npz differ diff --git a/bark/assets/prompts/v2/es_speaker_2.npz b/bark/assets/prompts/v2/es_speaker_2.npz new file mode 100644 index 0000000..ecbe04c Binary files /dev/null and b/bark/assets/prompts/v2/es_speaker_2.npz differ diff --git a/bark/assets/prompts/v2/es_speaker_3.npz b/bark/assets/prompts/v2/es_speaker_3.npz new file mode 100644 index 0000000..1d02fd7 Binary files /dev/null and b/bark/assets/prompts/v2/es_speaker_3.npz differ diff --git a/bark/assets/prompts/v2/es_speaker_4.npz b/bark/assets/prompts/v2/es_speaker_4.npz new file mode 100644 index 0000000..3881e0d Binary files /dev/null and b/bark/assets/prompts/v2/es_speaker_4.npz differ diff --git a/bark/assets/prompts/v2/es_speaker_5.npz b/bark/assets/prompts/v2/es_speaker_5.npz new file mode 100644 index 0000000..bf67639 Binary files /dev/null and b/bark/assets/prompts/v2/es_speaker_5.npz differ diff --git a/bark/assets/prompts/v2/es_speaker_6.npz b/bark/assets/prompts/v2/es_speaker_6.npz new file mode 100644 index 0000000..4c0b79f Binary files /dev/null and b/bark/assets/prompts/v2/es_speaker_6.npz differ diff --git a/bark/assets/prompts/v2/es_speaker_7.npz b/bark/assets/prompts/v2/es_speaker_7.npz new file mode 100644 index 0000000..d9bcde1 Binary files /dev/null and b/bark/assets/prompts/v2/es_speaker_7.npz differ diff --git a/bark/assets/prompts/v2/es_speaker_8.npz b/bark/assets/prompts/v2/es_speaker_8.npz new file mode 100644 index 0000000..ffc5483 Binary files /dev/null and b/bark/assets/prompts/v2/es_speaker_8.npz differ diff --git a/bark/assets/prompts/v2/es_speaker_9.npz b/bark/assets/prompts/v2/es_speaker_9.npz new file mode 100644 index 0000000..c68e682 Binary files /dev/null and b/bark/assets/prompts/v2/es_speaker_9.npz differ diff --git a/bark/assets/prompts/v2/fr_speaker_0.npz b/bark/assets/prompts/v2/fr_speaker_0.npz new file mode 100644 index 0000000..1868338 Binary files /dev/null and b/bark/assets/prompts/v2/fr_speaker_0.npz differ diff --git a/bark/assets/prompts/v2/fr_speaker_1.npz b/bark/assets/prompts/v2/fr_speaker_1.npz new file mode 100644 index 0000000..6b5a90b Binary files /dev/null and b/bark/assets/prompts/v2/fr_speaker_1.npz differ diff --git a/bark/assets/prompts/v2/fr_speaker_2.npz b/bark/assets/prompts/v2/fr_speaker_2.npz new file mode 100644 index 0000000..68a7df3 Binary files /dev/null and b/bark/assets/prompts/v2/fr_speaker_2.npz differ diff --git a/bark/assets/prompts/v2/fr_speaker_3.npz b/bark/assets/prompts/v2/fr_speaker_3.npz new file mode 100644 index 0000000..1c01916 Binary files /dev/null and b/bark/assets/prompts/v2/fr_speaker_3.npz differ diff --git a/bark/assets/prompts/v2/fr_speaker_4.npz b/bark/assets/prompts/v2/fr_speaker_4.npz new file mode 100644 index 0000000..ca1cdd7 Binary files /dev/null and b/bark/assets/prompts/v2/fr_speaker_4.npz differ diff --git a/bark/assets/prompts/v2/fr_speaker_5.npz b/bark/assets/prompts/v2/fr_speaker_5.npz new file mode 100644 index 0000000..5f20543 Binary files /dev/null and b/bark/assets/prompts/v2/fr_speaker_5.npz differ diff --git a/bark/assets/prompts/v2/fr_speaker_6.npz b/bark/assets/prompts/v2/fr_speaker_6.npz new file mode 100644 index 0000000..31e98c6 Binary files /dev/null and b/bark/assets/prompts/v2/fr_speaker_6.npz differ diff --git a/bark/assets/prompts/v2/fr_speaker_7.npz b/bark/assets/prompts/v2/fr_speaker_7.npz new file mode 100644 index 0000000..bcae0a2 Binary files /dev/null and b/bark/assets/prompts/v2/fr_speaker_7.npz differ diff --git a/bark/assets/prompts/v2/fr_speaker_8.npz b/bark/assets/prompts/v2/fr_speaker_8.npz new file mode 100644 index 0000000..baa1c08 Binary files /dev/null and b/bark/assets/prompts/v2/fr_speaker_8.npz differ diff --git a/bark/assets/prompts/v2/fr_speaker_9.npz b/bark/assets/prompts/v2/fr_speaker_9.npz new file mode 100644 index 0000000..2b1e251 Binary files /dev/null and b/bark/assets/prompts/v2/fr_speaker_9.npz differ diff --git a/bark/assets/prompts/v2/hi_speaker_0.npz b/bark/assets/prompts/v2/hi_speaker_0.npz new file mode 100644 index 0000000..c455fd7 Binary files /dev/null and b/bark/assets/prompts/v2/hi_speaker_0.npz differ diff --git a/bark/assets/prompts/v2/hi_speaker_1.npz b/bark/assets/prompts/v2/hi_speaker_1.npz new file mode 100644 index 0000000..1c3f19b Binary files /dev/null and b/bark/assets/prompts/v2/hi_speaker_1.npz differ diff --git a/bark/assets/prompts/v2/hi_speaker_2.npz b/bark/assets/prompts/v2/hi_speaker_2.npz new file mode 100644 index 0000000..7806586 Binary files /dev/null and b/bark/assets/prompts/v2/hi_speaker_2.npz differ diff --git a/bark/assets/prompts/v2/hi_speaker_3.npz b/bark/assets/prompts/v2/hi_speaker_3.npz new file mode 100644 index 0000000..2d74f31 Binary files /dev/null and b/bark/assets/prompts/v2/hi_speaker_3.npz differ diff --git a/bark/assets/prompts/v2/hi_speaker_4.npz b/bark/assets/prompts/v2/hi_speaker_4.npz new file mode 100644 index 0000000..46d877f Binary files /dev/null and b/bark/assets/prompts/v2/hi_speaker_4.npz differ diff --git a/bark/assets/prompts/v2/hi_speaker_5.npz b/bark/assets/prompts/v2/hi_speaker_5.npz new file mode 100644 index 0000000..1bf5c8d Binary files /dev/null and b/bark/assets/prompts/v2/hi_speaker_5.npz differ diff --git a/bark/assets/prompts/v2/hi_speaker_6.npz b/bark/assets/prompts/v2/hi_speaker_6.npz new file mode 100644 index 0000000..0e72810 Binary files /dev/null and b/bark/assets/prompts/v2/hi_speaker_6.npz differ diff --git a/bark/assets/prompts/v2/hi_speaker_7.npz b/bark/assets/prompts/v2/hi_speaker_7.npz new file mode 100644 index 0000000..2befa5a Binary files /dev/null and b/bark/assets/prompts/v2/hi_speaker_7.npz differ diff --git a/bark/assets/prompts/v2/hi_speaker_8.npz b/bark/assets/prompts/v2/hi_speaker_8.npz new file mode 100644 index 0000000..3d1c4e9 Binary files /dev/null and b/bark/assets/prompts/v2/hi_speaker_8.npz differ diff --git a/bark/assets/prompts/v2/hi_speaker_9.npz b/bark/assets/prompts/v2/hi_speaker_9.npz new file mode 100644 index 0000000..730698a Binary files /dev/null and b/bark/assets/prompts/v2/hi_speaker_9.npz differ diff --git a/bark/assets/prompts/v2/it_speaker_0.npz b/bark/assets/prompts/v2/it_speaker_0.npz new file mode 100644 index 0000000..c7132ba Binary files /dev/null and b/bark/assets/prompts/v2/it_speaker_0.npz differ diff --git a/bark/assets/prompts/v2/it_speaker_1.npz b/bark/assets/prompts/v2/it_speaker_1.npz new file mode 100644 index 0000000..cee09c9 Binary files /dev/null and b/bark/assets/prompts/v2/it_speaker_1.npz differ diff --git a/bark/assets/prompts/v2/it_speaker_2.npz b/bark/assets/prompts/v2/it_speaker_2.npz new file mode 100644 index 0000000..cae1fe3 Binary files /dev/null and b/bark/assets/prompts/v2/it_speaker_2.npz differ diff --git a/bark/assets/prompts/v2/it_speaker_3.npz b/bark/assets/prompts/v2/it_speaker_3.npz new file mode 100644 index 0000000..f7e970b Binary files /dev/null and b/bark/assets/prompts/v2/it_speaker_3.npz differ diff --git a/bark/assets/prompts/v2/it_speaker_4.npz b/bark/assets/prompts/v2/it_speaker_4.npz new file mode 100644 index 0000000..82888dd Binary files /dev/null and b/bark/assets/prompts/v2/it_speaker_4.npz differ diff --git a/bark/assets/prompts/v2/it_speaker_5.npz b/bark/assets/prompts/v2/it_speaker_5.npz new file mode 100644 index 0000000..c8cce48 Binary files /dev/null and b/bark/assets/prompts/v2/it_speaker_5.npz differ diff --git a/bark/assets/prompts/v2/it_speaker_6.npz b/bark/assets/prompts/v2/it_speaker_6.npz new file mode 100644 index 0000000..3b086b3 Binary files /dev/null and b/bark/assets/prompts/v2/it_speaker_6.npz differ diff --git a/bark/assets/prompts/v2/it_speaker_7.npz b/bark/assets/prompts/v2/it_speaker_7.npz new file mode 100644 index 0000000..16f4b43 Binary files /dev/null and b/bark/assets/prompts/v2/it_speaker_7.npz differ diff --git a/bark/assets/prompts/v2/it_speaker_8.npz b/bark/assets/prompts/v2/it_speaker_8.npz new file mode 100644 index 0000000..682aad2 Binary files /dev/null and b/bark/assets/prompts/v2/it_speaker_8.npz differ diff --git a/bark/assets/prompts/v2/it_speaker_9.npz b/bark/assets/prompts/v2/it_speaker_9.npz new file mode 100644 index 0000000..09c60c3 Binary files /dev/null and b/bark/assets/prompts/v2/it_speaker_9.npz differ diff --git a/bark/assets/prompts/v2/ja_speaker_0.npz b/bark/assets/prompts/v2/ja_speaker_0.npz new file mode 100644 index 0000000..d2a8d56 Binary files /dev/null and b/bark/assets/prompts/v2/ja_speaker_0.npz differ diff --git a/bark/assets/prompts/v2/ja_speaker_1.npz b/bark/assets/prompts/v2/ja_speaker_1.npz new file mode 100644 index 0000000..96016df Binary files /dev/null and b/bark/assets/prompts/v2/ja_speaker_1.npz differ diff --git a/bark/assets/prompts/v2/ja_speaker_2.npz b/bark/assets/prompts/v2/ja_speaker_2.npz new file mode 100644 index 0000000..3eeedc2 Binary files /dev/null and b/bark/assets/prompts/v2/ja_speaker_2.npz differ diff --git a/bark/assets/prompts/v2/ja_speaker_3.npz b/bark/assets/prompts/v2/ja_speaker_3.npz new file mode 100644 index 0000000..8037585 Binary files /dev/null and b/bark/assets/prompts/v2/ja_speaker_3.npz differ diff --git a/bark/assets/prompts/v2/ja_speaker_4.npz b/bark/assets/prompts/v2/ja_speaker_4.npz new file mode 100644 index 0000000..351a693 Binary files /dev/null and b/bark/assets/prompts/v2/ja_speaker_4.npz differ diff --git a/bark/assets/prompts/v2/ja_speaker_5.npz b/bark/assets/prompts/v2/ja_speaker_5.npz new file mode 100644 index 0000000..b4ab9e8 Binary files /dev/null and b/bark/assets/prompts/v2/ja_speaker_5.npz differ diff --git a/bark/assets/prompts/v2/ja_speaker_6.npz b/bark/assets/prompts/v2/ja_speaker_6.npz new file mode 100644 index 0000000..b56af0a Binary files /dev/null and b/bark/assets/prompts/v2/ja_speaker_6.npz differ diff --git a/bark/assets/prompts/v2/ja_speaker_7.npz b/bark/assets/prompts/v2/ja_speaker_7.npz new file mode 100644 index 0000000..74d4b46 Binary files /dev/null and b/bark/assets/prompts/v2/ja_speaker_7.npz differ diff --git a/bark/assets/prompts/v2/ja_speaker_8.npz b/bark/assets/prompts/v2/ja_speaker_8.npz new file mode 100644 index 0000000..3f2d224 Binary files /dev/null and b/bark/assets/prompts/v2/ja_speaker_8.npz differ diff --git a/bark/assets/prompts/v2/ja_speaker_9.npz b/bark/assets/prompts/v2/ja_speaker_9.npz new file mode 100644 index 0000000..1a5a545 Binary files /dev/null and b/bark/assets/prompts/v2/ja_speaker_9.npz differ diff --git a/bark/assets/prompts/v2/ko_speaker_0.npz b/bark/assets/prompts/v2/ko_speaker_0.npz new file mode 100644 index 0000000..1d4eeda Binary files /dev/null and b/bark/assets/prompts/v2/ko_speaker_0.npz differ diff --git a/bark/assets/prompts/v2/ko_speaker_1.npz b/bark/assets/prompts/v2/ko_speaker_1.npz new file mode 100644 index 0000000..90510ff Binary files /dev/null and b/bark/assets/prompts/v2/ko_speaker_1.npz differ diff --git a/bark/assets/prompts/v2/ko_speaker_2.npz b/bark/assets/prompts/v2/ko_speaker_2.npz new file mode 100644 index 0000000..adc8774 Binary files /dev/null and b/bark/assets/prompts/v2/ko_speaker_2.npz differ diff --git a/bark/assets/prompts/v2/ko_speaker_3.npz b/bark/assets/prompts/v2/ko_speaker_3.npz new file mode 100644 index 0000000..42d9690 Binary files /dev/null and b/bark/assets/prompts/v2/ko_speaker_3.npz differ diff --git a/bark/assets/prompts/v2/ko_speaker_4.npz b/bark/assets/prompts/v2/ko_speaker_4.npz new file mode 100644 index 0000000..a486d79 Binary files /dev/null and b/bark/assets/prompts/v2/ko_speaker_4.npz differ diff --git a/bark/assets/prompts/v2/ko_speaker_5.npz b/bark/assets/prompts/v2/ko_speaker_5.npz new file mode 100644 index 0000000..b3c6857 Binary files /dev/null and b/bark/assets/prompts/v2/ko_speaker_5.npz differ diff --git a/bark/assets/prompts/v2/ko_speaker_6.npz b/bark/assets/prompts/v2/ko_speaker_6.npz new file mode 100644 index 0000000..c966005 Binary files /dev/null and b/bark/assets/prompts/v2/ko_speaker_6.npz differ diff --git a/bark/assets/prompts/v2/ko_speaker_7.npz b/bark/assets/prompts/v2/ko_speaker_7.npz new file mode 100644 index 0000000..9098a97 Binary files /dev/null and b/bark/assets/prompts/v2/ko_speaker_7.npz differ diff --git a/bark/assets/prompts/v2/ko_speaker_8.npz b/bark/assets/prompts/v2/ko_speaker_8.npz new file mode 100644 index 0000000..e5eb9ff Binary files /dev/null and b/bark/assets/prompts/v2/ko_speaker_8.npz differ diff --git a/bark/assets/prompts/v2/ko_speaker_9.npz b/bark/assets/prompts/v2/ko_speaker_9.npz new file mode 100644 index 0000000..60086b0 Binary files /dev/null and b/bark/assets/prompts/v2/ko_speaker_9.npz differ diff --git a/bark/assets/prompts/v2/pl_speaker_0.npz b/bark/assets/prompts/v2/pl_speaker_0.npz new file mode 100644 index 0000000..03b8f5f Binary files /dev/null and b/bark/assets/prompts/v2/pl_speaker_0.npz differ diff --git a/bark/assets/prompts/v2/pl_speaker_1.npz b/bark/assets/prompts/v2/pl_speaker_1.npz new file mode 100644 index 0000000..a33e3ba Binary files /dev/null and b/bark/assets/prompts/v2/pl_speaker_1.npz differ diff --git a/bark/assets/prompts/v2/pl_speaker_2.npz b/bark/assets/prompts/v2/pl_speaker_2.npz new file mode 100644 index 0000000..b236952 Binary files /dev/null and b/bark/assets/prompts/v2/pl_speaker_2.npz differ diff --git a/bark/assets/prompts/v2/pl_speaker_3.npz b/bark/assets/prompts/v2/pl_speaker_3.npz new file mode 100644 index 0000000..04fab12 Binary files /dev/null and b/bark/assets/prompts/v2/pl_speaker_3.npz differ diff --git a/bark/assets/prompts/v2/pl_speaker_4.npz b/bark/assets/prompts/v2/pl_speaker_4.npz new file mode 100644 index 0000000..48dbf19 Binary files /dev/null and b/bark/assets/prompts/v2/pl_speaker_4.npz differ diff --git a/bark/assets/prompts/v2/pl_speaker_5.npz b/bark/assets/prompts/v2/pl_speaker_5.npz new file mode 100644 index 0000000..80fcbd9 Binary files /dev/null and b/bark/assets/prompts/v2/pl_speaker_5.npz differ diff --git a/bark/assets/prompts/v2/pl_speaker_6.npz b/bark/assets/prompts/v2/pl_speaker_6.npz new file mode 100644 index 0000000..46c86cd Binary files /dev/null and b/bark/assets/prompts/v2/pl_speaker_6.npz differ diff --git a/bark/assets/prompts/v2/pl_speaker_7.npz b/bark/assets/prompts/v2/pl_speaker_7.npz new file mode 100644 index 0000000..49a7f20 Binary files /dev/null and b/bark/assets/prompts/v2/pl_speaker_7.npz differ diff --git a/bark/assets/prompts/v2/pl_speaker_8.npz b/bark/assets/prompts/v2/pl_speaker_8.npz new file mode 100644 index 0000000..8bb593d Binary files /dev/null and b/bark/assets/prompts/v2/pl_speaker_8.npz differ diff --git a/bark/assets/prompts/v2/pl_speaker_9.npz b/bark/assets/prompts/v2/pl_speaker_9.npz new file mode 100644 index 0000000..7031a57 Binary files /dev/null and b/bark/assets/prompts/v2/pl_speaker_9.npz differ diff --git a/bark/assets/prompts/v2/pt_speaker_0.npz b/bark/assets/prompts/v2/pt_speaker_0.npz new file mode 100644 index 0000000..76aa02a Binary files /dev/null and b/bark/assets/prompts/v2/pt_speaker_0.npz differ diff --git a/bark/assets/prompts/v2/pt_speaker_1.npz b/bark/assets/prompts/v2/pt_speaker_1.npz new file mode 100644 index 0000000..0119298 Binary files /dev/null and b/bark/assets/prompts/v2/pt_speaker_1.npz differ diff --git a/bark/assets/prompts/v2/pt_speaker_2.npz b/bark/assets/prompts/v2/pt_speaker_2.npz new file mode 100644 index 0000000..9e4e401 Binary files /dev/null and b/bark/assets/prompts/v2/pt_speaker_2.npz differ diff --git a/bark/assets/prompts/v2/pt_speaker_3.npz b/bark/assets/prompts/v2/pt_speaker_3.npz new file mode 100644 index 0000000..2198310 Binary files /dev/null and b/bark/assets/prompts/v2/pt_speaker_3.npz differ diff --git a/bark/assets/prompts/v2/pt_speaker_4.npz b/bark/assets/prompts/v2/pt_speaker_4.npz new file mode 100644 index 0000000..21c8f0e Binary files /dev/null and b/bark/assets/prompts/v2/pt_speaker_4.npz differ diff --git a/bark/assets/prompts/v2/pt_speaker_5.npz b/bark/assets/prompts/v2/pt_speaker_5.npz new file mode 100644 index 0000000..3245b43 Binary files /dev/null and b/bark/assets/prompts/v2/pt_speaker_5.npz differ diff --git a/bark/assets/prompts/v2/pt_speaker_6.npz b/bark/assets/prompts/v2/pt_speaker_6.npz new file mode 100644 index 0000000..f705295 Binary files /dev/null and b/bark/assets/prompts/v2/pt_speaker_6.npz differ diff --git a/bark/assets/prompts/v2/pt_speaker_7.npz b/bark/assets/prompts/v2/pt_speaker_7.npz new file mode 100644 index 0000000..c60affd Binary files /dev/null and b/bark/assets/prompts/v2/pt_speaker_7.npz differ diff --git a/bark/assets/prompts/v2/pt_speaker_8.npz b/bark/assets/prompts/v2/pt_speaker_8.npz new file mode 100644 index 0000000..64c9d8c Binary files /dev/null and b/bark/assets/prompts/v2/pt_speaker_8.npz differ diff --git a/bark/assets/prompts/v2/pt_speaker_9.npz b/bark/assets/prompts/v2/pt_speaker_9.npz new file mode 100644 index 0000000..c7b891b Binary files /dev/null and b/bark/assets/prompts/v2/pt_speaker_9.npz differ diff --git a/bark/assets/prompts/v2/ru_speaker_0.npz b/bark/assets/prompts/v2/ru_speaker_0.npz new file mode 100644 index 0000000..4af14b5 Binary files /dev/null and b/bark/assets/prompts/v2/ru_speaker_0.npz differ diff --git a/bark/assets/prompts/v2/ru_speaker_1.npz b/bark/assets/prompts/v2/ru_speaker_1.npz new file mode 100644 index 0000000..8ab3c7e Binary files /dev/null and b/bark/assets/prompts/v2/ru_speaker_1.npz differ diff --git a/bark/assets/prompts/v2/ru_speaker_2.npz b/bark/assets/prompts/v2/ru_speaker_2.npz new file mode 100644 index 0000000..2bdfc39 Binary files /dev/null and b/bark/assets/prompts/v2/ru_speaker_2.npz differ diff --git a/bark/assets/prompts/v2/ru_speaker_3.npz b/bark/assets/prompts/v2/ru_speaker_3.npz new file mode 100644 index 0000000..32047d1 Binary files /dev/null and b/bark/assets/prompts/v2/ru_speaker_3.npz differ diff --git a/bark/assets/prompts/v2/ru_speaker_4.npz b/bark/assets/prompts/v2/ru_speaker_4.npz new file mode 100644 index 0000000..5c3f47a Binary files /dev/null and b/bark/assets/prompts/v2/ru_speaker_4.npz differ diff --git a/bark/assets/prompts/v2/ru_speaker_5.npz b/bark/assets/prompts/v2/ru_speaker_5.npz new file mode 100644 index 0000000..6c73ed3 Binary files /dev/null and b/bark/assets/prompts/v2/ru_speaker_5.npz differ diff --git a/bark/assets/prompts/v2/ru_speaker_6.npz b/bark/assets/prompts/v2/ru_speaker_6.npz new file mode 100644 index 0000000..6171e1d Binary files /dev/null and b/bark/assets/prompts/v2/ru_speaker_6.npz differ diff --git a/bark/assets/prompts/v2/ru_speaker_7.npz b/bark/assets/prompts/v2/ru_speaker_7.npz new file mode 100644 index 0000000..294c34d Binary files /dev/null and b/bark/assets/prompts/v2/ru_speaker_7.npz differ diff --git a/bark/assets/prompts/v2/ru_speaker_8.npz b/bark/assets/prompts/v2/ru_speaker_8.npz new file mode 100644 index 0000000..abd95d8 Binary files /dev/null and b/bark/assets/prompts/v2/ru_speaker_8.npz differ diff --git a/bark/assets/prompts/v2/ru_speaker_9.npz b/bark/assets/prompts/v2/ru_speaker_9.npz new file mode 100644 index 0000000..06ab868 Binary files /dev/null and b/bark/assets/prompts/v2/ru_speaker_9.npz differ diff --git a/bark/assets/prompts/v2/tr_speaker_0.npz b/bark/assets/prompts/v2/tr_speaker_0.npz new file mode 100644 index 0000000..6d6be17 Binary files /dev/null and b/bark/assets/prompts/v2/tr_speaker_0.npz differ diff --git a/bark/assets/prompts/v2/tr_speaker_1.npz b/bark/assets/prompts/v2/tr_speaker_1.npz new file mode 100644 index 0000000..76a6f8d Binary files /dev/null and b/bark/assets/prompts/v2/tr_speaker_1.npz differ diff --git a/bark/assets/prompts/v2/tr_speaker_2.npz b/bark/assets/prompts/v2/tr_speaker_2.npz new file mode 100644 index 0000000..09b86b1 Binary files /dev/null and b/bark/assets/prompts/v2/tr_speaker_2.npz differ diff --git a/bark/assets/prompts/v2/tr_speaker_3.npz b/bark/assets/prompts/v2/tr_speaker_3.npz new file mode 100644 index 0000000..4e64e64 Binary files /dev/null and b/bark/assets/prompts/v2/tr_speaker_3.npz differ diff --git a/bark/assets/prompts/v2/tr_speaker_4.npz b/bark/assets/prompts/v2/tr_speaker_4.npz new file mode 100644 index 0000000..2e238f6 Binary files /dev/null and b/bark/assets/prompts/v2/tr_speaker_4.npz differ diff --git a/bark/assets/prompts/v2/tr_speaker_5.npz b/bark/assets/prompts/v2/tr_speaker_5.npz new file mode 100644 index 0000000..ad62887 Binary files /dev/null and b/bark/assets/prompts/v2/tr_speaker_5.npz differ diff --git a/bark/assets/prompts/v2/tr_speaker_6.npz b/bark/assets/prompts/v2/tr_speaker_6.npz new file mode 100644 index 0000000..586d29f Binary files /dev/null and b/bark/assets/prompts/v2/tr_speaker_6.npz differ diff --git a/bark/assets/prompts/v2/tr_speaker_7.npz b/bark/assets/prompts/v2/tr_speaker_7.npz new file mode 100644 index 0000000..64a8c61 Binary files /dev/null and b/bark/assets/prompts/v2/tr_speaker_7.npz differ diff --git a/bark/assets/prompts/v2/tr_speaker_8.npz b/bark/assets/prompts/v2/tr_speaker_8.npz new file mode 100644 index 0000000..f5a50cd Binary files /dev/null and b/bark/assets/prompts/v2/tr_speaker_8.npz differ diff --git a/bark/assets/prompts/v2/tr_speaker_9.npz b/bark/assets/prompts/v2/tr_speaker_9.npz new file mode 100644 index 0000000..6c1b234 Binary files /dev/null and b/bark/assets/prompts/v2/tr_speaker_9.npz differ diff --git a/bark/assets/prompts/v2/zh_speaker_0.npz b/bark/assets/prompts/v2/zh_speaker_0.npz new file mode 100644 index 0000000..0bd0db1 Binary files /dev/null and b/bark/assets/prompts/v2/zh_speaker_0.npz differ diff --git a/bark/assets/prompts/v2/zh_speaker_1.npz b/bark/assets/prompts/v2/zh_speaker_1.npz new file mode 100644 index 0000000..8c593ef Binary files /dev/null and b/bark/assets/prompts/v2/zh_speaker_1.npz differ diff --git a/bark/assets/prompts/v2/zh_speaker_2.npz b/bark/assets/prompts/v2/zh_speaker_2.npz new file mode 100644 index 0000000..eda3b26 Binary files /dev/null and b/bark/assets/prompts/v2/zh_speaker_2.npz differ diff --git a/bark/assets/prompts/v2/zh_speaker_3.npz b/bark/assets/prompts/v2/zh_speaker_3.npz new file mode 100644 index 0000000..d410990 Binary files /dev/null and b/bark/assets/prompts/v2/zh_speaker_3.npz differ diff --git a/bark/assets/prompts/v2/zh_speaker_4.npz b/bark/assets/prompts/v2/zh_speaker_4.npz new file mode 100644 index 0000000..a0417e8 Binary files /dev/null and b/bark/assets/prompts/v2/zh_speaker_4.npz differ diff --git a/bark/assets/prompts/v2/zh_speaker_5.npz b/bark/assets/prompts/v2/zh_speaker_5.npz new file mode 100644 index 0000000..e24abe3 Binary files /dev/null and b/bark/assets/prompts/v2/zh_speaker_5.npz differ diff --git a/bark/assets/prompts/v2/zh_speaker_6.npz b/bark/assets/prompts/v2/zh_speaker_6.npz new file mode 100644 index 0000000..efc598a Binary files /dev/null and b/bark/assets/prompts/v2/zh_speaker_6.npz differ diff --git a/bark/assets/prompts/v2/zh_speaker_7.npz b/bark/assets/prompts/v2/zh_speaker_7.npz new file mode 100644 index 0000000..fdecb43 Binary files /dev/null and b/bark/assets/prompts/v2/zh_speaker_7.npz differ diff --git a/bark/assets/prompts/v2/zh_speaker_8.npz b/bark/assets/prompts/v2/zh_speaker_8.npz new file mode 100644 index 0000000..0b775c5 Binary files /dev/null and b/bark/assets/prompts/v2/zh_speaker_8.npz differ diff --git a/bark/assets/prompts/v2/zh_speaker_9.npz b/bark/assets/prompts/v2/zh_speaker_9.npz new file mode 100644 index 0000000..04fdfae Binary files /dev/null and b/bark/assets/prompts/v2/zh_speaker_9.npz differ diff --git a/bark/generation.py b/bark/generation.py index 64b3c47..3769734 100644 --- a/bark/generation.py +++ b/bark/generation.py @@ -1,9 +1,7 @@ import contextlib import gc -import hashlib import os import re -import requests from encodec import EncodecModel import funcy @@ -72,8 +70,9 @@ SUPPORTED_LANGS = [ ALLOWED_PROMPTS = {"announcer"} for _, lang in SUPPORTED_LANGS: - for n in range(10): - ALLOWED_PROMPTS.add(f"{lang}_speaker_{n}") + for prefix in ("", f"v2{os.path.sep}"): + for n in range(10): + ALLOWED_PROMPTS.add(f"{prefix}{lang}_speaker_{n}") logger = logging.getLogger(__name__) @@ -95,32 +94,26 @@ REMOTE_MODEL_PATHS = { "text_small": { "repo_id": "suno/bark", "file_name": "text.pt", - "checksum": "b3e42bcbab23b688355cd44128c4cdd3", }, "coarse_small": { "repo_id": "suno/bark", "file_name": "coarse.pt", - "checksum": "5fe964825e3b0321f9d5f3857b89194d", }, "fine_small": { "repo_id": "suno/bark", "file_name": "fine.pt", - "checksum": "5428d1befe05be2ba32195496e58dc90", }, "text": { "repo_id": "suno/bark", "file_name": "text_2.pt", - "checksum": "54afa89d65e318d4f5f80e8e8799026a", }, "coarse": { "repo_id": "suno/bark", "file_name": "coarse_2.pt", - "checksum": "8a98094e5e3a255a5c9c0ab7efe8fd28", }, "fine": { "repo_id": "suno/bark", "file_name": "fine_2.pt", - "checksum": "59d184ed44e3650774a2f0503a48a97b", }, } @@ -132,26 +125,6 @@ if not hasattr(torch.nn.functional, 'scaled_dot_product_attention') and torch.cu ) -def _string_md5(s): - m = hashlib.md5() - m.update(s.encode("utf-8")) - return m.hexdigest() - - -def _md5(fname): - hash_md5 = hashlib.md5() - with open(fname, "rb") as f: - for chunk in iter(lambda: f.read(4096), b""): - hash_md5.update(chunk) - return hash_md5.hexdigest() - - -def _get_ckpt_path(model_type, use_small=False): - model_key = f"{model_type}_small" if use_small or USE_SMALL_MODELS else model_type - model_name = _string_md5(REMOTE_MODEL_PATHS[model_key]["file_name"]) - return os.path.join(CACHE_DIR, f"{model_name}.pt") - - def _grab_best_device(use_gpu=True): if torch.cuda.device_count() > 0 and use_gpu: device = "cuda" @@ -162,11 +135,17 @@ def _grab_best_device(use_gpu=True): return device -def _download(from_hf_path, file_name, to_local_path): +def _get_ckpt_path(model_type, use_small=False): + key = model_type + if use_small: + key += "_small" + return os.path.join(CACHE_DIR, REMOTE_MODEL_PATHS[key]["file_name"]) + + +def _download(from_hf_path, file_name): os.makedirs(CACHE_DIR, exist_ok=True) - destination_file_name = to_local_path.split("/")[-1] hf_hub_download(repo_id=from_hf_path, filename=file_name, local_dir=CACHE_DIR) - os.replace(os.path.join(CACHE_DIR, file_name), to_local_path) + class InferenceContext: def __init__(self, benchmark=False): @@ -223,15 +202,9 @@ def _load_model(ckpt_path, device, use_small=False, model_type="text"): raise NotImplementedError() model_key = f"{model_type}_small" if use_small or USE_SMALL_MODELS else model_type model_info = REMOTE_MODEL_PATHS[model_key] - if ( - os.path.exists(ckpt_path) and - _md5(ckpt_path) != model_info["checksum"] - ): - logger.warning(f"found outdated {model_type} model, removing.") - os.remove(ckpt_path) if not os.path.exists(ckpt_path): logger.info(f"{model_type} model not found, downloading into `{CACHE_DIR}`.") - _download(model_info["repo_id"], model_info["file_name"], ckpt_path) + _download(model_info["repo_id"], model_info["file_name"]) checkpoint = torch.load(ckpt_path, map_location=device) # this is a hack model_args = checkpoint["model_args"] @@ -376,6 +349,25 @@ TEXT_PAD_TOKEN = 129_595 SEMANTIC_INFER_TOKEN = 129_599 +def _load_history_prompt(history_prompt_input): + if isinstance(history_prompt_input, str) and history_prompt_input.endswith(".npz"): + history_prompt = np.load(history_prompt_input) + elif isinstance(history_prompt_input, str): + if history_prompt_input not in ALLOWED_PROMPTS: + raise ValueError("history prompt not found") + history_prompt = np.load( + os.path.join(CUR_PATH, "assets", "prompts", f"{history_prompt_input}.npz") + ) + elif isinstance(history_prompt_input, dict): + assert("semantic_prompt" in history_prompt_input) + assert("coarse_prompt" in history_prompt_input) + assert("fine_prompt" in history_prompt_input) + history_prompt = history_prompt_input + else: + raise ValueError("history prompt format unrecognized") + return history_prompt + + def generate_text_semantic( text, history_prompt=None, @@ -393,13 +385,8 @@ def generate_text_semantic( text = _normalize_whitespace(text) assert len(text.strip()) > 0 if history_prompt is not None: - if history_prompt.endswith(".npz"): - semantic_history = np.load(history_prompt)["semantic_prompt"] - else: - assert (history_prompt in ALLOWED_PROMPTS) - semantic_history = np.load( - os.path.join(CUR_PATH, "assets", "prompts", f"{history_prompt}.npz") - )["semantic_prompt"] + history_prompt = _load_history_prompt(history_prompt) + semantic_history = history_prompt["semantic_prompt"] assert ( isinstance(semantic_history, np.ndarray) and len(semantic_history.shape) == 1 @@ -562,15 +549,9 @@ def generate_coarse( semantic_to_coarse_ratio = COARSE_RATE_HZ / SEMANTIC_RATE_HZ * N_COARSE_CODEBOOKS max_semantic_history = int(np.floor(max_coarse_history / semantic_to_coarse_ratio)) if history_prompt is not None: - if history_prompt.endswith(".npz"): - x_history = np.load(history_prompt) - else: - assert (history_prompt in ALLOWED_PROMPTS) - x_history = np.load( - os.path.join(CUR_PATH, "assets", "prompts", f"{history_prompt}.npz") - ) - x_semantic_history = x_history["semantic_prompt"] - x_coarse_history = x_history["coarse_prompt"] + history_prompt = _load_history_prompt(history_prompt) + x_semantic_history = history_prompt["semantic_prompt"] + x_coarse_history = history_prompt["coarse_prompt"] assert ( isinstance(x_semantic_history, np.ndarray) and len(x_semantic_history.shape) == 1 @@ -727,13 +708,8 @@ def generate_fine( and x_coarse_gen.max() <= CODEBOOK_SIZE - 1 ) if history_prompt is not None: - if history_prompt.endswith(".npz"): - x_fine_history = np.load(history_prompt)["fine_prompt"] - else: - assert (history_prompt in ALLOWED_PROMPTS) - x_fine_history = np.load( - os.path.join(CUR_PATH, "assets", "prompts", f"{history_prompt}.npz") - )["fine_prompt"] + history_prompt = _load_history_prompt(history_prompt) + x_fine_history = history_prompt["fine_prompt"] assert ( isinstance(x_fine_history, np.ndarray) and len(x_fine_history.shape) == 2 diff --git a/notebooks/long_form_generation.ipynb b/notebooks/long_form_generation.ipynb new file mode 100644 index 0000000..e4ce61e --- /dev/null +++ b/notebooks/long_form_generation.ipynb @@ -0,0 +1,400 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": 1, + "id": "39ea4bed", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "\n", + "os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"0\"\n", + "\n", + "\n", + "from IPython.display import Audio\n", + "import nltk # we'll use this to split into sentences\n", + "import numpy as np\n", + "\n", + "from bark.generation import (\n", + " generate_text_semantic,\n", + " preload_models,\n", + ")\n", + "from bark.api import semantic_to_waveform\n", + "from bark import generate_audio, SAMPLE_RATE" + ] + }, + { + "cell_type": "code", + "execution_count": 29, + "id": "776964b6", + "metadata": {}, + "outputs": [], + "source": [ + "preload_models()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "1d03f4d2", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "markdown", + "id": "74a025a4", + "metadata": {}, + "source": [ + "# Simple Long-Form Generation\n", + "We split longer text into sentences using `nltk` and generate the sentences one by one." + ] + }, + { + "cell_type": "code", + "execution_count": 33, + "id": "57b06e2a", + "metadata": {}, + "outputs": [], + "source": [ + "script = \"\"\"\n", + "Hey, have you heard about this new text-to-audio model called \"Bark\"? \n", + "Apparently, it's the most realistic and natural-sounding text-to-audio model \n", + "out there right now. People are saying it sounds just like a real person speaking. \n", + "I think it uses advanced machine learning algorithms to analyze and understand the \n", + "nuances of human speech, and then replicates those nuances in its own speech output. \n", + "It's pretty impressive, and I bet it could be used for things like audiobooks or podcasts. \n", + "In fact, I heard that some publishers are already starting to use Bark to create audiobooks. \n", + "It would be like having your own personal voiceover artist. I really think Bark is going to \n", + "be a game-changer in the world of text-to-audio technology.\n", + "\"\"\".replace(\"\\n\", \" \").strip()" + ] + }, + { + "cell_type": "code", + "execution_count": 34, + "id": "f747f804", + "metadata": {}, + "outputs": [], + "source": [ + "sentences = nltk.sent_tokenize(script)" + ] + }, + { + "cell_type": "code", + "execution_count": 35, + "id": "17400a9b", + "metadata": { + "scrolled": true + }, + "outputs": [ + { + "name": "stderr", + "output_type": "stream", + "text": [ + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 100/100 [00:02<00:00, 43.03it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 17/17 [00:06<00:00, 2.45it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 100/100 [00:04<00:00, 22.73it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 33/33 [00:13<00:00, 2.52it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 100/100 [00:01<00:00, 66.30it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 11/11 [00:04<00:00, 2.46it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 100/100 [00:04<00:00, 20.99it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 35/35 [00:14<00:00, 2.46it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 100/100 [00:03<00:00, 25.63it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 29/29 [00:11<00:00, 2.50it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 100/100 [00:04<00:00, 23.90it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 30/30 [00:12<00:00, 2.46it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 100/100 [00:01<00:00, 53.24it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 14/14 [00:05<00:00, 2.51it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 100/100 [00:01<00:00, 50.63it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 15/15 [00:05<00:00, 2.57it/s]\n" + ] + } + ], + "source": [ + "SPEAKER = \"v2/en_speaker_6\"\n", + "silence = np.zeros(int(0.25 * SAMPLE_RATE)) # quarter second of silence\n", + "\n", + "pieces = []\n", + "for sentence in sentences:\n", + " audio_array = generate_audio(sentence, history_prompt=SPEAKER)\n", + " pieces += [audio_array, silence.copy()]\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "04cf77f9", + "metadata": {}, + "outputs": [], + "source": [ + "Audio(np.concatenate(pieces), rate=SAMPLE_RATE)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ac2d4625", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "markdown", + "id": "6d13249b", + "metadata": {}, + "source": [ + "# $ \\\\ $" + ] + }, + { + "cell_type": "markdown", + "id": "cdfc8bf5", + "metadata": {}, + "source": [ + "# Advanced Long-Form Generation\n", + "Somtimes Bark will hallucinate a little extra audio at the end of the prompt.\n", + "We can solve this issue by lowering the threshold for bark to stop generating text. \n", + "We use the `min_eos_p` kwarg in `generate_text_semantic`" + ] + }, + { + "cell_type": "code", + "execution_count": 37, + "id": "62807fd0", + "metadata": {}, + "outputs": [ + { + "name": "stderr", + "output_type": "stream", + "text": [ + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 100/100 [00:02<00:00, 38.05it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 18/18 [00:07<00:00, 2.46it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 100/100 [00:03<00:00, 32.28it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 21/21 [00:08<00:00, 2.54it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 100/100 [00:01<00:00, 55.78it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 14/14 [00:05<00:00, 2.57it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 100/100 [00:06<00:00, 14.73it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 35/35 [00:14<00:00, 2.47it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 100/100 [00:02<00:00, 40.29it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 18/18 [00:07<00:00, 2.56it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 100/100 [00:03<00:00, 32.92it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 20/20 [00:08<00:00, 2.47it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 100/100 [00:01<00:00, 68.87it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 12/12 [00:04<00:00, 2.62it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 100/100 [00:02<00:00, 47.64it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 15/15 [00:06<00:00, 2.46it/s]\n" + ] + } + ], + "source": [ + "GEN_TEMP = 0.6\n", + "SPEAKER = \"v2/en_speaker_6\"\n", + "silence = np.zeros(int(0.25 * SAMPLE_RATE)) # quarter second of silence\n", + "\n", + "pieces = []\n", + "for sentence in sentences:\n", + " semantic_tokens = generate_text_semantic(\n", + " sentence,\n", + " history_prompt=SPEAKER,\n", + " temp=GEN_TEMP,\n", + " min_eos_p=0.05, # this controls how likely the generation is to end\n", + " )\n", + "\n", + " audio_array = semantic_to_waveform(semantic_tokens, history_prompt=SPEAKER,)\n", + " pieces += [audio_array, silence.copy()]\n", + "\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "133fec46", + "metadata": {}, + "outputs": [], + "source": [ + "Audio(np.concatenate(pieces), rate=SAMPLE_RATE)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6eee9f5a", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "markdown", + "id": "be8e125e", + "metadata": {}, + "source": [ + "# $ \\\\ $" + ] + }, + { + "cell_type": "markdown", + "id": "03a16c1b", + "metadata": {}, + "source": [ + "# Make a Long-Form Dialog with Bark" + ] + }, + { + "cell_type": "markdown", + "id": "06c5eff8", + "metadata": {}, + "source": [ + "### Step 1: Format a script and speaker lookup" + ] + }, + { + "cell_type": "code", + "execution_count": 14, + "id": "5238b297", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "['Samantha: Hey, have you heard about this new text-to-audio model called \"Bark\"?',\n", + " \"John: No, I haven't. What's so special about it?\",\n", + " \"Samantha: Well, apparently it's the most realistic and natural-sounding text-to-audio model out there right now. People are saying it sounds just like a real person speaking.\",\n", + " 'John: Wow, that sounds amazing. How does it work?',\n", + " 'Samantha: I think it uses advanced machine learning algorithms to analyze and understand the nuances of human speech, and then replicates those nuances in its own speech output.',\n", + " \"John: That's pretty impressive. Do you think it could be used for things like audiobooks or podcasts?\",\n", + " 'Samantha: Definitely! In fact, I heard that some publishers are already starting to use Bark to create audiobooks. And I bet it would be great for podcasts too.',\n", + " 'John: I can imagine. It would be like having your own personal voiceover artist.',\n", + " 'Samantha: Exactly! I think Bark is going to be a game-changer in the world of text-to-audio technology.']" + ] + }, + "execution_count": 14, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "speaker_lookup = {\"Samantha\": \"v2/en_speaker_9\", \"John\": \"v2/en_speaker_2\"}\n", + "\n", + "# Script generated by chat GPT\n", + "script = \"\"\"\n", + "Samantha: Hey, have you heard about this new text-to-audio model called \"Bark\"?\n", + "\n", + "John: No, I haven't. What's so special about it?\n", + "\n", + "Samantha: Well, apparently it's the most realistic and natural-sounding text-to-audio model out there right now. People are saying it sounds just like a real person speaking.\n", + "\n", + "John: Wow, that sounds amazing. How does it work?\n", + "\n", + "Samantha: I think it uses advanced machine learning algorithms to analyze and understand the nuances of human speech, and then replicates those nuances in its own speech output.\n", + "\n", + "John: That's pretty impressive. Do you think it could be used for things like audiobooks or podcasts?\n", + "\n", + "Samantha: Definitely! In fact, I heard that some publishers are already starting to use Bark to create audiobooks. And I bet it would be great for podcasts too.\n", + "\n", + "John: I can imagine. It would be like having your own personal voiceover artist.\n", + "\n", + "Samantha: Exactly! I think Bark is going to be a game-changer in the world of text-to-audio technology.\"\"\"\n", + "script = script.strip().split(\"\\n\")\n", + "script = [s.strip() for s in script if s]\n", + "script" + ] + }, + { + "cell_type": "markdown", + "id": "ee547efd", + "metadata": {}, + "source": [ + "### Step 2: Generate the audio for every speaker turn" + ] + }, + { + "cell_type": "code", + "execution_count": 15, + "id": "203e5081", + "metadata": {}, + "outputs": [ + { + "name": "stderr", + "output_type": "stream", + "text": [ + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 100/100 [00:02<00:00, 34.03it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 22/22 [00:08<00:00, 2.55it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 100/100 [00:01<00:00, 71.58it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 11/11 [00:04<00:00, 2.65it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 100/100 [00:04<00:00, 22.75it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 33/33 [00:13<00:00, 2.53it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 100/100 [00:01<00:00, 70.76it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 11/11 [00:04<00:00, 2.63it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 100/100 [00:04<00:00, 20.46it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 36/36 [00:14<00:00, 2.47it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 100/100 [00:04<00:00, 20.18it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 37/37 [00:14<00:00, 2.51it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 100/100 [00:04<00:00, 23.04it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 32/32 [00:12<00:00, 2.48it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 100/100 [00:01<00:00, 54.64it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 14/14 [00:05<00:00, 2.58it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 100/100 [00:03<00:00, 31.71it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 24/24 [00:09<00:00, 2.56it/s]\n" + ] + } + ], + "source": [ + "pieces = []\n", + "silence = np.zeros(int(0.5*SAMPLE_RATE))\n", + "for line in script:\n", + " speaker, text = line.split(\": \")\n", + " audio_array = generate_audio(text, history_prompt=speaker_lookup[speaker], )\n", + " pieces += [audio_array, silence.copy()]" + ] + }, + { + "cell_type": "markdown", + "id": "7c54bada", + "metadata": {}, + "source": [ + "### Step 3: Concatenate all of the audio and play it" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "27a56842", + "metadata": {}, + "outputs": [], + "source": [ + "Audio(np.concatenate(pieces), rate=SAMPLE_RATE)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a1bc5877", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.9.16" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/notebooks/memory_profiling_bark.ipynb b/notebooks/memory_profiling_bark.ipynb new file mode 100644 index 0000000..4c500da --- /dev/null +++ b/notebooks/memory_profiling_bark.ipynb @@ -0,0 +1,187 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "90641144", + "metadata": {}, + "source": [ + "# Bark Memory Profiling\n", + "Bark has two ways to reduce GPU memory: \n", + " - Small models: a smaller version of the model. This can be set by using the environment variable `SUNO_USE_SMALL_MODELS`\n", + " - offloading models to CPU: Holding only one model at a time on the GPU, and shuttling the models to the CPU in between generations. \n", + "\n", + "## NOTE: this requires a GPU to run\n", + "\n", + "# $ \\\\ $\n", + "## First, we'll use the most memory efficient configuration" + ] + }, + { + "cell_type": "code", + "execution_count": 1, + "id": "39ea4bed", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "\n", + "os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"0\"\n", + "os.environ[\"SUNO_USE_SMALL_MODELS\"] = \"1\"\n", + "os.environ[\"SUNO_OFFLOAD_CPU\"] = \"1\"\n", + "\n", + "from bark.generation import preload_models\n", + "from bark import generate_audio, SAMPLE_RATE\n", + "\n", + "import torch" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "id": "66b0c006", + "metadata": {}, + "outputs": [ + { + "name": "stderr", + "output_type": "stream", + "text": [ + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 100/100 [00:01<00:00, 62.17it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 10/10 [00:03<00:00, 2.74it/s]\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "max memory usage = 2396MB\n" + ] + } + ], + "source": [ + "torch.cuda.reset_peak_memory_stats()\n", + "preload_models()\n", + "audio_array = generate_audio(\"madam I'm adam\", history_prompt=\"v2/en_speaker_5\")\n", + "max_utilization = torch.cuda.max_memory_allocated()\n", + "print(f\"max memory usage = {max_utilization / 1024 / 1024:.0f}MB\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9922dd2d", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "bdbe578e", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "markdown", + "id": "213d1b5b", + "metadata": {}, + "source": [ + "# Memory Profiling:\n", + "We can profile the memory consumption of 4 scenarios\n", + " - Small models, offloading to CPU\n", + " - Large models, offloading to CPU\n", + " - Small models, not offloading to CPU\n", + " - Large models, not offloading to CPU" + ] + }, + { + "cell_type": "code", + "execution_count": 1, + "id": "417d5e9c", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "from bark.generation import preload_models\n", + "from bark import generate_audio, SAMPLE_RATE\n", + "import torch\n", + "import time" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "id": "cd83b45d", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Small models True, offloading to CPU: True\n", + "\tmax memory usage = 2949MB, time 3s\n", + "\n", + "Small models False, offloading to CPU: True\n", + "\tmax memory usage = 7826MB, time 4s\n", + "\n", + "Small models True, offloading to CPU: False\n", + "\tmax memory usage = 5504MB, time 2s\n", + "\n", + "Small models False, offloading to CPU: False\n", + "\tmax memory usage = 7825MB, time 5s\n", + "\n" + ] + } + ], + "source": [ + "offload_models = True\n", + "use_small_models = True\n", + "\n", + "for offload_models in (True, False):\n", + " for use_small_models in (True, False):\n", + " torch.cuda.reset_peak_memory_stats()\n", + " preload_models(\n", + " text_use_small=use_small_models,\n", + " coarse_use_small=use_small_models,\n", + " fine_use_small=use_small_models,\n", + " force_reload=True,\n", + " )\n", + " t0 = time.time()\n", + " audio_array = generate_audio(\"madam I'm adam\", history_prompt=\"v2/en_speaker_5\", silent=True)\n", + " dur = time.time() - t0\n", + " max_utilization = torch.cuda.max_memory_allocated()\n", + " print(f\"Small models {use_small_models}, offloading to CPU: {offload_models}\")\n", + " print(f\"\\tmax memory usage = {max_utilization / 1024 / 1024:.0f}MB, time {dur:.0f}s\\n\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "bfe5fa06", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.9.16" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/notebooks/use_small_models_on_cpu.ipynb b/notebooks/use_small_models_on_cpu.ipynb new file mode 100644 index 0000000..4ec79c5 --- /dev/null +++ b/notebooks/use_small_models_on_cpu.ipynb @@ -0,0 +1,142 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "6a682b61", + "metadata": {}, + "source": [ + "# Benchmarking small models on CPU\n", + " - We can enable small models with the `SUNO_USE_SMALL_MODELS` environment variable" + ] + }, + { + "cell_type": "code", + "execution_count": 1, + "id": "9500dd93", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "\n", + "os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"\"\n", + "os.environ[\"SUNO_USE_SMALL_MODELS\"] = \"1\"\n", + "\n", + "from IPython.display import Audio\n", + "import numpy as np\n", + "\n", + "from bark import generate_audio, preload_models, SAMPLE_RATE\n", + "\n", + "import time" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "id": "4e3454b6", + "metadata": {}, + "outputs": [ + { + "name": "stderr", + "output_type": "stream", + "text": [ + "No GPU being used. Careful, inference might be very slow!\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "CPU times: user 5.52 s, sys: 2.34 s, total: 7.86 s\n", + "Wall time: 4.33 s\n" + ] + } + ], + "source": [ + "%%time\n", + "preload_models()" + ] + }, + { + "cell_type": "code", + "execution_count": 3, + "id": "f6024e5f", + "metadata": {}, + "outputs": [ + { + "name": "stderr", + "output_type": "stream", + "text": [ + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 100/100 [00:10<00:00, 9.89it/s]\n", + "100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 15/15 [00:43<00:00, 2.90s/it]\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "took 62s to generate 6s of audio\n" + ] + } + ], + "source": [ + "t0 = time.time()\n", + "text = \"In the light of the moon, a little egg lay on a leaf\"\n", + "audio_array = generate_audio(text)\n", + "generation_duration_s = time.time() - t0\n", + "audio_duration_s = audio_array.shape[0] / SAMPLE_RATE\n", + "\n", + "print(f\"took {generation_duration_s:.0f}s to generate {audio_duration_s:.0f}s of audio\")" + ] + }, + { + "cell_type": "code", + "execution_count": 4, + "id": "2dcce86c", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "10" + ] + }, + "execution_count": 4, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "os.cpu_count()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3046eddb", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.9.16" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +}