Open czhou opened 1 year ago
Not yet. It's on my list of things to do. It'll likely be awhile since I've been extremely busy.
cli will be great!!!
Yeah, I would also appreciate a command line feature !
Working on a bash script right now, which downloads any video, splits the stems and transcribes lyrics. As spleeter is not really usable anymore, being able to split stems via cli would be a huge improvement!
Not yet. It's on my list of things to do. It'll likely be awhile since I've been extremely busy.
Any progress information?
Hey folks, in case this is helpful to anyone...
I have a simple karaoke creation use case and I already know the MDX models trained by @Anjok07 work for UVR work really well for this (especially UVR_MDXNET_KARA_2
), but I wanted to be able to easily run the separation from the command line or from my other python hobby projects.
So, I created this little python package, splitting out the minimal parts from the UVR repo in order to do a simple separation of a WAV file into the primary & secondary stems WAV files.
If your use case is simple or similar to mine (e.g. you don't need any non-MDX models, you don't need more than 2 stems, or secondary processing, advanced settings, etc. - you just need a song to be split into instrumental and vocals - then this may be useful to you!
It can either be installed using pip and run from the CLI, or imported as a python package and used in some other project (example in README).
https://github.com/karaokenerds/python-audio-separator
To use on the CLI, you can simply install it from PyPI:
pip install audio-separator
and run with a single CLI param specifying a WAV file to separate, e.g.
audio-separator input.wav
and it should output 2 WAV files in the current directory after about 20 seconds.
it defaults to using the UVR_MDXNET_KARA_2
model, but you can use any other model just by specifying it by name with the --model_name
CLI arg/param.
It downloads both the model metadata and model file itself on demand to /tmp/audio-separator-models
, but you can also override this with the --model_file_dir
arg/param.
To be clear: what I've created here is really just a refactoring and simplification of one part of Anjok's code - he absolutely deserves credit if this is useful to anyone! (https://www.buymeacoffee.com/uvr5 👀)
and uh, I hope this is ok @Anjok07 😅 - I did consider forking UVR and trying to raise a PR here to add a CLI mode, but I kinda wanted a super minimal thing which just worked without any of the additional dependencies etc. in the full-fat UVR repo.
Oh, and contributions / PRs very much welcome of course!
@beveradb Bro thanks, I was just making myself a bot in tg and wanted to add this functionality.
A simple CLI version of UVR: https://github.com/seanghay/uvr
@Anjok07 Any chance of CLI happening soon?
like:
uvr --model=htdemucs_ft --model=another_model --input=input.mp3 --output-dir=/path/to/ouput