Local Text To Speech feature
Quick setup guide with coqui-tts
The installation process can be difficult for non-developers. Also, the github repo will probably change a lot in the coming weeks.
To make the experience more immersive, you can enable the TTS feature by following this installation guide.
Thanks to the help of a user from the community, here's a lightweight TTS python script you can run on your computer to synthesize voices!
I've rushed the integration in eudaimonia to make sure it would be available fast. But I intend to develop it further.
1. Miniconda:
You must first download and install miniconda to create a virtual environment.
Download Miniconda
2. Virtual env:
Once miniconda has been installed, open the miniconda terminal
Create then activate your virtual environment:
conda create -n local-tts python=3.10 -y
conda activate local-tts
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118
3. Python server:
Download or clone the local-TTS repo on github.
If you encounter an 'espeak' error, you need to install espeak-ng
Follow the installation instructions here: .
Then restart your computer for the environment Path to be updated.
In the miniconda terminal, move inside the local-tts folder you've just downloaded:
cd "YOUR/LOCAL-TTS/PATH"
Install the dependencies:
pip install -r requirements.txt
Then start the server on the port you want:
uvicorn server:app --host 0.0.0.0 --port 3200
Once you see this in your terminal, it means TTS is ready.
4. Eudaimonia:
You can now fill the TTS Server url in eudaimonia with: http://localhost:3200
For now you can choose between two voices. More voices will be available later, and probably with voice cloning. The first TTS request can take time since you'll need to download and load the model.
- CPU: fast and light can run on the cpu alone
- GPU: Slightly slower and requiring ~4GB VRAM (from the GPU), but better quality and voice cloning capability
Don't forget that everytime you want to use the TTS feature, you'll have to start miniconda and activate your virtual env:
conda activate local-tts
uvicorn server:app --host 0.0.0.0 --port 3200
4. Runpod:
You can also use runpod for the TTS feature.
Follow the installation steps here.
Here is a short video guide to help you setup the TTS server on Runpod: