Renaming files within audio-separator

[Bebra777228]

I suggest improving the functionality of the audio-separator by adding two new parameters to the separate function: primary_name and secondary_name.

As a result, the code will look like this:

INPUT_PATH = "/content/input/music.mp3"
OUTPUT_PATH = "/content/output"
VOCALS = os.path.join(OUTPUT_PATH, "Vocals.wav")
INSTRUMENTAL = os.path.join(OUTPUT_PATH, "Instrumental.wav")

separator = Separator(output_dir=OUTPUT_PATH)
separator.load_model(model_filename="model_bs_roformer_ep_317_sdr_12.9755.ckpt")

voc_inst_result = separator.separate(INPUT_PATH, VOCALS, INSTRUMENTAL)

# or
voc_inst_result = separator.separate(INPUT_PATH, primary_name=VOCALS, secondary_name=INSTRUMENTAL)

If these parameters are not specified, the files will be output in the same way they are saved now.

This will be a more convenient option compared to the current approach, which requires additional file renaming:

INPUT_PATH = "/content/input/music.mp3"
OUTPUT_PATH = "/content/output"
VOCALS = os.path.join(OUTPUT_PATH, "Vocals.wav")
INSTRUMENTAL = os.path.join(OUTPUT_PATH, "Instrumental.wav")

separator = Separator(output_dir=OUTPUT_PATH)
separator.load_model(model_filename="model_bs_roformer_ep_317_sdr_12.9755.ckpt")

voc_inst_result = separator.separate(INPUT_PATH)
os.rename(os.path.join(output_dir, voc_inst_result[0]), INSTRUMENTAL)
os.rename(os.path.join(output_dir, voc_inst_result[1]), VOCALS)

Hope my idea is clear!

[beveradb]

Sure, fine by me if it doesn't break existing workflows - if you implement this in a PR, please add tests :)

[Bebra777228]

I don't think I can do it, but I'll try :)

[StarMariner]

Q. Would the os.rename solve a problem with output-dir giving filenames with parenthesis around the stem names?
I am unable to mv or cp the filenames in an Ubuntu dash Shell script. I don't know where the parenthesis are added to the path.

audio-separator "$file" --output_dir "$output_dir" --output_format wav --single_stem vocals
Gives an output postfix of: _(Vocals)model_mel_band_roformer_ep_3005_sdr_1.wav
Filename such as : noisey_story(Vocals)_model_mel_band_roformer_ep_3005_sdr_1.wav

[beveradb]

PR #141 now merged, so the new CLI parameters --primary_output_name and --secondary_output_name are available from audio-separator version 0.24 and above.

Thanks for your contribution @Bebra777228 🙇

and @StarMariner this is probably one option to help you out now, since you could specify simple output filenames with no special characters in them, e.g.
audio-separator -m UVR_MDXNET_KARA_2.onnx --primary_output_name instrumental --secondary_output_name vocals input.wav

However, if you open a discussion thread and share your shell script code, I can probably help you learn how to handle special characters properly - that's basic shell scripting stuff which you can learn on the internet pretty easily (especially with the help of LLMs these days) so you should be able to get it working 😄