coqui-ai
/
TTS
mirror of https://github.com/coqui-ai/TTS.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Update Bark Config

pull/2685/head
Eren G??lge 2023-06-21 11:58:18 +02:00
parent 695e862aad
commit 03c347b7f3
2 changed files with 38 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
TTS/bin/synthesize.py
View File

@ -356,7 +356,7 @@ If you don't specify any models, then it uses LJSpeech based English model.
vc_config_path = config_path
# tts model with multiple files to be loaded from the directory path
if isinstance(model_item["github_rls_url"], list):
if isinstance(model_item["model_url"], list):
model_dir = model_path
tts_path = None
tts_config_path = None

39
TTS/tts/configs/bark_config.py
View File

@ -1,5 +1,5 @@
import os
from dataclasses import dataclass, field
from dataclasses import dataclass
from typing import Dict
from TTS.tts.configs.shared_configs import BaseTTSConfig
@ -11,6 +11,40 @@ from TTS.utils.generic_utils import get_user_data_dir
@dataclass
class BarkConfig(BaseTTSConfig):
""" Bark TTS configuration
Args:
model (str): model name that registers the model.
audio (BarkAudioConfig): audio configuration. Defaults to BarkAudioConfig().
num_chars (int): number of characters in the alphabet. Defaults to 0.
semantic_config (GPTConfig): semantic configuration. Defaults to GPTConfig().
fine_config (FineGPTConfig): fine configuration. Defaults to FineGPTConfig().
coarse_config (GPTConfig): coarse configuration. Defaults to GPTConfig().
CONTEXT_WINDOW_SIZE (int): GPT context window size. Defaults to 1024.
SEMANTIC_RATE_HZ (float): semantic tokens rate in Hz. Defaults to 49.9.
SEMANTIC_VOCAB_SIZE (int): semantic vocabulary size. Defaults to 10_000.
CODEBOOK_SIZE (int): encodec codebook size. Defaults to 1024.
N_COARSE_CODEBOOKS (int): number of coarse codebooks. Defaults to 2.
N_FINE_CODEBOOKS (int): number of fine codebooks. Defaults to 8.
COARSE_RATE_HZ (int): coarse tokens rate in Hz. Defaults to 75.
SAMPLE_RATE (int): sample rate. Defaults to 24_000.
USE_SMALLER_MODELS (bool): use smaller models. Defaults to False.
TEXT_ENCODING_OFFSET (int): text encoding offset. Defaults to 10_048.
SEMANTIC_PAD_TOKEN (int): semantic pad token. Defaults to 10_000.
TEXT_PAD_TOKEN ([type]): text pad token. Defaults to 10_048.
TEXT_EOS_TOKEN ([type]): text end of sentence token. Defaults to 10_049.
TEXT_SOS_TOKEN ([type]): text start of sentence token. Defaults to 10_050.
SEMANTIC_INFER_TOKEN (int): semantic infer token. Defaults to 10_051.
COARSE_SEMANTIC_PAD_TOKEN (int): coarse semantic pad token. Defaults to 12_048.
COARSE_INFER_TOKEN (int): coarse infer token. Defaults to 12_050.
REMOTE_BASE_URL ([type]): remote base url. Defaults to "https://huggingface.co/erogol/bark/tree".
REMOTE_MODEL_PATHS (Dict): remote model paths. Defaults to None.
LOCAL_MODEL_PATHS (Dict): local model paths. Defaults to None.
SMALL_REMOTE_MODEL_PATHS (Dict): small remote model paths. Defaults to None.
CACHE_DIR (str): local cache directory. Defaults to get_user_data_dir().
DEF_SPEAKER_DIR (str): default speaker directory to stoke speaker values for voice cloning. Defaults to get_user_data_dir().
"""
model: str = "bark"
audio: BarkAudioConfig = BarkAudioConfig()
num_chars: int = 0
@ -39,6 +73,7 @@ class BarkConfig(BaseTTSConfig):
LOCAL_MODEL_PATHS: Dict = None
SMALL_REMOTE_MODEL_PATHS: Dict = None
CACHE_DIR: str = str(get_user_data_dir("tts/suno/bark_v0"))
DEF_SPEAKER_DIR: str = str(get_user_data_dir("tts/bark_v0/speakers"))
def __post_init__(self):
self.REMOTE_MODEL_PATHS = {
@ -67,4 +102,4 @@ class BarkConfig(BaseTTSConfig):
"coarse": {"path": os.path.join(self.REMOTE_BASE_URL, "coarse.pt")},
"fine": {"path": os.path.join(self.REMOTE_BASE_URL, "fine.pt")},
}
self.sample_rate = self.SAMPLE_RATE
self.sample_rate = self.SAMPLE_RATE # pylint: disable=attribute-defined-outside-init