Automate dataset creation

Haurrus / xtts-trainer-no-ui-auto

This project includes a Python script for fine-tuning a text-to-speech (TTS) model. The script utilizes custom datasets and use CUDA for accelerated training.

13 stars 4 forks source link

Automate dataset creation #1

Open platform-kit opened 10 months ago

platform-kit commented 10 months ago

Hi, this script looks very useful.

Does it automate dataset creation too? If not, I think that would be a great feature to add, so that one can simply point it at a directory of audio of various formats (mp3 / wav at least) and then chop them into clips, and generate the .CSV.

Haurrus commented 10 months ago

Yeah I could add this, it was already in the web-ui repo. It was using fast_whisper to do it, I could add it.

platform-kit commented 10 months ago

That would be amazing!

platform-kit commented 9 months ago

Hey @Haurrus any progress on this?

Haurrus commented 9 months ago

I added this xtts_generate_dataset.py to create dataset, you can look into the readme :

Audio Dataset Preprocessing xtts_generate_dataset.py