Info

This fork contains a completely reworked Streamlit GUI. The 152334H fork of Tortoise-TTS has the best likeness to imported voices at the moment, in my humble opinion. The mrq version has much better nuances and control but adds an American accent to most of my imported voices that are not fine tuned which was driving me nuts. I started working on this because I just wanted to have the extra settings mrq had on their Gradio UI. So I taught myself some Streamlit and python and got the GUI working. I've also fixed a few annoying issues along the way and added some extra features. I'm not sure how far I can develop this or keep it updated with my knowledge but for the time being it works.

Jump To

• Features
• Screenshots
• FAQ
• Prerequisites
• Installaton
  • Run GUI
  • Troubleshooting
    • Poetry Missing
    • Modules Missing
• To Do
• How To Make Proper Samples For Cloning

Features

• Fully reworked GUI
• Fully control all generation parameters.
• Generate and dump conditioning latent for faster generations or for use with other ML models.
• Download multiple generated candidates.

Screenshots

FAQ

1. Will this work on an older 8gb and under card? Yes it will. Just check Low VRAM in the gui.
2. Is the install easy? It relatively is. There is a proper (AHEM!) step by step guide below.
3. Can I save conditional latents? Yes you can!

Prerequisites

1. CudaToolkit : Download from here and install. Choose custom install and uncheck everything except the cudatoolkit option. Your graphic driver will have more updated drivers for the other components you DON'T WANT TO CHANGE THAT.
2. Visual C + : Download from here and install. Make sure you select Use in Python

Installaton

To install on Windows please follow these steps (It was very unclear in the original repo and a lot of users had issues) : Open a miniconda/anaconda terminal as an administrator

conda create --name ttsgui python==3.8

conda activate ttsgui

If you have Tortoise installed before next run :

pip uninstall tortoise

Install Cuda 11 to your environment

conda install cudatoolkit=11.7

Next navigate to the drive/folder where you want the tortoise-fast-tts folder to be extracted to. Then run :

git clone https://github.com/Acephalia/tortoise-tts-fast-GUI
cd tortoise-tts-fast-GUI

conda install pytorch==1.13.1 torchvision==0.14.1 torchaudio==0.13.1 pytorch-cuda=11.7 -c pytorch -c nvidia

pip install -e .

pip install git+https://github.com/152334H/BigVGAN.git

Run GUI

To Run open anaconda/miniconda terminal :

conda activate ttsgui
s:
cd tortoise-tts-fast-GUI
streamlit run app.py

Set your drive letter and path in my case s:/tortoise-tts-fast-GUI

Troubleshooting

Poetry Missing

Install Poetry into your environment

conda activate ttsgui
s:
cd tortoise-tts-fast-GUI
pip install poetry 
poetry install
poetry shell

Set your drive letter and path in my case s:/tortoise-tts-fast-GUI

Modules Missing

If you encounter any messages stating xyz module is missing or the likes simply go back into your env and run

pip install xyz

Replace xyz and repeat till all missing modules are installed.

To Do :

1. Add remaining settings
2. Fix files being overwritten in results folder
3. Fix Download Audio file
4. Fix multiple candidate download
5. Add voice refresh
6. Add GUI to create and download conditional latents
7. Automatically create conditional latents in voice folder

How To Make Proper Samples For Cloning

Coming Soon

---

Visit the original repo for more info.