This bot provides the ability to synthesize voice samples using tts-tortoise API from local host. It has multilanguage support and voice-to-voice conversion via whisper model.
There should be a running instance of the bot, if you would like to check it out at https://t.me/tts_tbot
Here's the preview:
- all of the features of tts-tortoise, including voice cloning, emotion steering, multisampling
- user-settings database
- custom voice adding
- voice-to-voice conversion via mic recording
- NVIDIA GPU (between 4 and 10Gb of VRAM required for inference, depending on tortoise settings)
- python 3.10.6
- pip or anaconda env with python 3.10.6
- ffmpeg
Tested on Linux only, written with cross-platform in mind, should work on Windows. Although tortoise-tts environment may fail to resolve.
Clone the repo with submodules
git clone --recurse-submodules $(URL)
For example, let's create new conda environment and install bot there
conda create -n voice_tbot python=3.10.6
conda activate voice_tbot
cd to repo directory and execute to install dependencies:
python -m pip install -r ./requirements.txt
Create configuration file named "config" inside bot_data directory and set user parameters. Use this example as reference.
Run module package voice_bot from the required environment, for example - from repo directory:
python -m voice_bot
Use punctuation (ellipses, exclamation points, CAPS, semicolons, commas) to add emphasis and shape the speech. You can also try to add different emotions to sentences by prepending parts of texts with "[describe emotion]" notations (For more info please visit the original model source page).
- add synthensis from audio files
This project is licensed under the MIT License - see the LICENSE file for details.