Added script for ljspeech creation from mrs sqlite database

README.md

@@ -56,8 +56,14 @@ Contributions are accepted! We'd love the communities help in building a better
 ### Training
 
 *Note: you need at least 40GB of free disk space to train a model.*
+0. **Use mimic-recording-studio data as ljspeech dataset.**
+  * Edit your local environment parameters in createljspeech.py
+  * Run createljspeech.py
+  * Copy ljspeech file structure to `~/tacotron`
+> Go forward to step 2
 
 1. **Download a speech dataset.**
+> If using your own mimic-recording-studio dataset please use step 0 insted
 
    The following are supported out of the box:
     * [LJ Speech](https://keithito.com/LJ-Speech-Dataset/) (Public Domain)

createljspeech.py

@@ -0,0 +1,54 @@
+# This script generates a folder structure for ljspeech-1.1 processing from mimic-recording-studio database
+# Written as first python script (eg. helloworld) by Thorsten Mueller (MrThorstenM@gmx.net) on november 2019 without any warranty
+
+import sqlite3
+import os
+from datetime import datetime
+from shutil import copyfile
+
+# please set following vars matching your environment
+# ---------------------------------------------------
+
+# Directory of mimic-recording-studio (eg: /home/thorsten/mimic-recording-studio)
+directory_base_mrs = "/home/thorsten/tts/mimic-recording-studio"
+# Output directory for ljspeech structure (eg: /tmp/tts)
+directory_base_ljspeech = "/home/thorsten/tts"
+# See /mimic-recording-studio/backend/audio_files/<id> (eg: 3beeae88-0777-2c8c-5c93-2e844a462e26)
+speaker_id = "4aeeae88-0777-2c8c-5c93-2e844a462e49" 
+
+# -----------------------
+# end of environment vars
+
+# Create needed folder structure for ljspeech-1.1
+def folder_creation():
+    now = datetime.now()
+    dt_string = now.strftime("%d.%m.%Y_%H-%M-%S")
+
+    global directory_base_ljspeech
+    directory_base_ljspeech = directory_base_ljspeech + "/ljspeech_" + dt_string + "/LJSpeech-1.1"
+
+    if not os.path.exists(directory_base_ljspeech):
+                os.makedirs(directory_base_ljspeech)
+
+    if not os.path.exists(directory_base_ljspeech + "/wavs"):
+                os.makedirs(directory_base_ljspeech + "/wavs")
+
+def main():
+    folder_creation()
+
+    conn = sqlite3.connect(directory_base_mrs + '/backend/db/mimicstudio.db')
+    c = conn.cursor()
+
+    # Create new metadata.csv for ljspeech
+    metadata = open(directory_base_ljspeech + "/metadata.csv",mode="w", encoding="utf8") 
+
+    # todo: Check best order for performance of tacotron training
+    for row in c.execute('SELECT audio_id, prompt, lower(prompt) FROM audiomodel ORDER BY length(prompt)'):
+        metadata.write(row[0] + "|" + row[1] + "|" + row[2] + "\n")
+        copyfile(directory_base_mrs + "/backend/audio_files/" + speaker_id + "/" + row[0] + ".wav",directory_base_ljspeech + "/wavs/" + row[0] + ".wav")
+
+    metadata.close()
+    conn.close()
+
+if __name__ == '__main__':
+    main()