Batch file processing

[lenover12]

I have trained a model on my voice and I want to batch convert thousands of small audio files to make a game mod. Is there a way to do this?

assuming I need to use the CLI, can I choose a pre-set I created?

[lenover12]

I found a solution with the help and annoyance of chatGPT
my goal was to find every .ogg file in a parent and its child directories and process them with the voice model, outputting them with their directory path in so-vits-svc-fork\infer_output\ + the directory path

Create a .bat file like this for use on windows 10: (change the paths as you want, and the arguments)
cd D:\so-vits-svc-fork SETLOCAL ENABLEDELAYEDEXPANSION FOR /R "C:\Users\user\Desktop\parent_directory" %%f IN (*.ogg) DO ( set "rel_path=%%~pf" set "rel_path=!rel_path:~14,-1!" set "out_path=D:\so-vits-svc-fork\infer_output\!rel_path!\%%~nf.ogg" MD "!out_path:\%%~nxf=!" 2>NUL svc infer "%%f" --output-path "!out_path!" --speaker "lenover" --f0-method "crepe" --transpose -15 --pad-seconds 0.1 )

then in your virtual environment in command prompt (as administrator) run the .bat file

[nikich340]

That's not a real solution, inferencing every files separately requires model to be loaded/unloaded to gpu every time.

[lenover12]

before batch processing was added it helped me convert thousands of files, so If anyone else needed a quick and dirty solution I shared what I got working.

[GarrettConway]

Wow, I'm amazed that worked out for you. I'll take a first pass at adding in a simple CLI batch convert tonight so hopefully this isn't as painful in the future.

[nikich340]

Currently GUI supports selecting a folder for inference, what equals to batch processing.