captain-flam / karafan Goto Github PK

Describe the bug
I found it impossible to process files with more than a model using the slowest speed

To Reproduce
Steps to reproduce the behavior:

1. Go to vocal separation
2. select one model and select the second one next to it or from another column like music filter and select the slowest speed mode
3. Scroll down to start
4. See hanging and files won't be produced in Google Colab even after a time

Expected behavior
the files should be produced in Google Colab

Desktop (please complete the following information):

• OS: [Windows]
• Browser [Chrome]
• Version [Latest Stable]

Additional context
Add any other context about the problem here.

Hello!

First of all, congratulations on the application and all the documentation, it is perfectly excellent.

I think it would be awesome if KaraFan automatically copied song metadata like artist, album, title, synced lyrics to the generated outputs. What would also be cool is if the naming scheme of the output files could be changed to FILENAME_(vocals)

Why is this you may ask?

I'm using karaoke software that automatically downloads .lrc from a music file or from the Internet.
in the current state of the naming scheme the metadata is not being copied, this often fails and does not find the letter until the file is renamed accordingly.

I don't know how to program in Python to send a PR and to solve my problem I'm using the package node-id3 in my custom script to copy all the tags, This takes a little work as there are 3 steps (split, rename and copy).

All the best

Describe the bug
FileNotFoundError : [Errno 2] No such file or directory: '/Users/user/projects/ai/voice/KaraFan_user/Config_PC.ini'

I'm on OSX, but I can't find any where in the docs or the code how to generate this Config_PC.ini? The wiki and docs don't talk about it. Do I have to run the windows .exe file to generate this stuff? Is the python code not self contained?

Hi, I'd like to join your server. Could I get an updated invite please? The one on the GitHub page doesnt work anymore, thank you!

it does not start
File "C:\Users\hellraiza\Downloads\KaraFan-master\KaraFan-master\KaraFan.pyw", line 65, in
frame = Gui.wx_Main.KaraFanForm(None, {'Gdrive': Gdrive, 'Project': Project, 'isColab': False})
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "C:\Users\hellraiza\Downloads\KaraFan-master\KaraFan-master\Gui\wx_Main.py", line 52, in init
with open(os.path.join(params['Project'], "App", "init.py"), "r") as version_file:
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

please help

Code is becoming obsolete in new versions of Python and Google colab
The problems are slowness and incompatibility in Google colab with the GPU

Hello! Thank you so much for this beautiful code base, I have enjoyed separating out so many songs with it. I saw you had listed for an API, it would be awesome to have one on replicate or some other cloud provider.

I would love to be able to run this through an API so I can do it without a GPU or complicated setup.

Have you looked much into which GPU API service you are wanting to use, and if not, is there anything I could to do to help get that up?

Thanks Captain!

Colab stucks at connecting..... and progress does not update, and final stems are not produced

When running KaraFan 5.1 there is a pop-up command terminal instance every second that prevents me from navigating windows explorer or other applications. Seems to be a new behavior after the update introduced a real-time progress bar.

There is a possibility that Microsoft Visual C++ 2015-2019 (x64) is the issue. I could not install this older version from the Setup.py so I removed it from the script to finish the install. Otherwise there's no issues separating tracks, but the pop-ups make KaraFan 5.1 unusable for me otherwise.

Hey! is there a way to use this like I can use demucs, just by running it in the command line?

When trying to install KaraFan on an Apple Silicon Mac (specifically an M2 Ultra Mac Studio), I've run into multiple issues along the way. Most of which I've been able to fix, but the following error is something I can't get past. Terminal produces this error message when trying to run 'Setup.py':
ERROR: Could not find a version that satisfies the requirement onnxruntime-gpu (from versions: none) ERROR: No matching distribution found for onnxruntime-gpu

So far, my understanding is that onnxruntime-gpu cannot run on MacOS, and the inbuilt ML cores must be used instead of Nvidia's CUDA solution. But of course my technical knowledge doesn't go very deep.

To Reproduce
Steps to reproduce the behavior:

1. Install Homebrew, Python, etc. beforehand
2. Change 'Setup.py' to the correct Python version
3. Run 'Setup.py' in terminal (as double clicking just opens up an editor, it does not run the app)
4. See error

Expected behavior
Of course, the expected behaviour is that it will install correctly

Desktop (please complete the following information):

• OS: MacOS 13.5.1
• Machine: M2 Ultra Mac Studio

Describe the bug
On the progress tab, when the output text reaches the view limit, the app crashes. I found that the issue lies in the inference.py -> def write -> self.CONSOLE.ScrollLines(1). I believe that when it reaches the end of the screen, it tries to scroll the page, but the scroll component doesn't exist yet and it crashes.

To Reproduce
Steps to reproduce the behavior:

1. Go to the 'Settings' tab. (select some audio files)
2. Click on 'Start'.
3. Navigate to the 'Progress' tab.
4. Wait until the console text reaches the end of the screen.

Expected behavior
The app will crash.

Screenshots

Desktop (please complete the following information):

• OS: Windows 11
• Browser: Brave
• Version 1.65.132

Additional context
You can reduce the window height to speed up the error.

It is being extremely slow when choosing a non-MDX23C model as if It is using only CPU. RAM usage there is high instead of GPU usage and It stucks on it for long time!

It would be great if one could select between GPU or CPU processing, just like it is in UVR.
In my case, my system has a powerfull CPU but only a small passive graphics cards.
AFAIK there are recompiled version of torch for CPU and GPU available.

happens whenever i press start