This is an audiobook creator using Tortoise TTS

With this repo you will be able to generate super high quality audio books using ai models locally on your computer for absolutely FREE. (No internet connection is needed)

Sample Audio

Click below to listen to the audio sample:

This repo is a fork of the tortoise-fast repo: https://github.com/152334H/tortoise-tts-fast.git which was created from the repo: https://github.com/neonbjb/tortoise-tts.git

BIG THANKS TO THE ORIGINAL CREATOR OF TORTOISE AND THE CREATOR OF TORTOISE FAST!!!

I changed quite a few things:

• How text is split into chunks
• Added the ability to add pauses to generation. Single or double line breaks create short pause and more than 2 lines breaks a long pause that can be configured from streamlit gui
• Changed the UI quite a bit to optimize it for audio book creation
• Ability to load in a file
• Added self correction feature that tries to fix issues with generation automatically (word, char differences, and pitch control)
• Finetuned tortoise settings
• Change terminal logs to show generation details more clearly
• Saves each generated file as mp3 and preserves them upon new generation
• Changed the way as audio files were saved avoiding any quality loss
• Save/reset settings

Hardware used for testing

Nvidia RTX 3090 (with Cuda 11.7) Nvidia RTX 4090 (with Cuda 11.8)

Installation

I only tested it on Ubuntu 22.04 Linux.

Here are the steps:

1. Install latest proprietary nvidia driver
2. Install Ubuntu packages
   sudo apt install git git-lfs perl make ffmpeg nvidia-cuda-toolkit nvidia-cudnn libportaudio2
3. Download Miniconda from: https://docs.conda.io/projects/miniconda/en/latest/miniconda-install.html
4. Install it without sudo rights for current user
5. Restart computer
6. Clone this repo
   git clone https://github.com/georgecsaszargit/tortoise_audio_book_creator.git
7. CD into the repo folder where you can see the requirements-new.txt
8. Create conda env:
   conda env create -f environment-new.yml
9. Activate conda: 
   conda activate tortoiseaudiobook
10. Install python packages using pip:
    python -m pip install -r requirements-rtx3090.txt
11. Install tortoise module:
    python -m pip install -e .
12. ONLY ON RTX4090 do this 1 following line:
    pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118
13. Download finetuned models and place them to ~/.cache/tortoise/models/ folder from: https://huggingface.co/csdzs/tortoise-audiobook-creator-finetuned-models(These models are better than the original tortoise models)
    git clone https://huggingface.co/csdzs/tortoise-audiobook-creator-finetuned-models
    cd tortoise-audiobook-creator-finetuned-models
    git lfs fetch --all
    git lfs checkout
    mkdir -p ~/.cache/tortoise/models
    cp * ~/.cache/tortoise/models
14. cd 1 level up and run tortoise:
    cd ..
    streamlit run scripts/app.py

Instructional video: https://youtu.be/BCCMB0p4fC8?si=5pHqHb8nZCSa_ExO