Taein-TTS

Description

Table of Contents

• Installation
• Usage
• Contributing
• License

Installation

1. Clone the huggingface repository: https://huggingface.co/icecream0910/taein-tts
2. Modify the run-server.bat batch file in the /server directory to match your actual file paths.
   For example, if your server folder is at C:\myown-tts\server, update the file as follows:
   bat

   1@echo off 2setlocal 3cd /D "%~dp0" 4set MECAB_KO_DIC_PATH=.\mecab\mecab-ko-dic -r .\mecab\mecabrc 5set TTS_MODEL_FILE=C:\myown-tts\server\models\glowtts-v2\best_model.pth.tar 6set TTS_MODEL_CONFIG=C:\myown-tts\server\models\glowtts-v2\config.json 7set VOCODER_MODEL_FILE=C:\myown-tts\server\models\hifigan-v2\best_model.pth.tar 8set VOCODER_MODEL_CONFIG=C:\myown-tts\server\models\hifigan-v2\config.json 9server.exe 10endlocal
3. Update the glowtts-v2/config.json and hifigan-v2/config.json files in the /server/models/ directory with your actual file paths.
   Ensure you double the backslash (\\) in the file paths, as shown below:
   • For glowtts-v2/config.json:
   "stats_path": "C:\\mydata\\tts-server\\models\\glowtts-v2\\scale_stats.npy"
   • For hifigan-v2/config.json:
   "stats_path": "C:\\mydata\\tts-server\\models\\hifigan-v2\\scale_stats.npy"

Usage

API

• Text preprocessing: /tts-server/api/process-text
  Splits sentences and removes special characters to automatically stitch together and playback multi-line sentences as you type.
• Text Inference: /tts-server/api/infer-glowtts
  Synthesizes text to speech. Send the text to be synthesized in the text parameter of the URL.
  Example:
  http://localhost:5000/tts-server/api/infer-glowtts?text=hello

Text Inference Demo Page

Contributing

1. Fork the repository (https://github.com/icecream0910/myown-tts/fork).
2. Create a new branch: git checkout -b feature/<featureName>.
3. Commit your changes: git commit -am 'Add <featureName>'.
4. Push to the branch: git push origin feature/<featureName>.
5. Submit a pull request.

License

References

• SCE-TTS
• g2pK
• mimic-recording-studio
• coqui TTS

• Korean Corpus for Voice Recording
• SleepingCE Speech Dataset
• Pre-trained Models for SleepingCE Speech Dataset (Glow-TTS)
• Pre-trained Models for SleepingCE Speech Dataset (HiFi-GAN)
  • These models were fine-tuned from the model provided by coqui-ai/TTS, trained on the VCTK dataset, available here.