frikallo / misst

A local GUI music source separation tool built on Tkinter and demucs serving as a free and open source Stem Player

---

| MISST on Windows 11 with Dark mode and 'Blue' theme with 'Steve Lacy's Infrunami' playing

| MISST on Windows 11 with Light mode and 'Blue' theme with 'Frank Ocean's Ivy' playing

| MISST on Windows 11 Showcasing how versatile and personal you can be with MISST!

| MISST on Windows 11 Showcasing how importing audios is as easy as two clicks!

Original Repository of MISST : Music/Instrumental Stem Separation Tool.

This application uses state-of-the-art demucs source separation models to extract the 4 core stems from audio files (Bass, Drums, Other Instrumentals and Vocals). But it is not limited to this. MISST acts as a developped music player aswell, fit to enjoy and medal with your audio files as you see fit. MISST even comes prepared to import songs and playlists directly from your music library.

This project is OpenSource, feel free to use, study and/or send pull request.

Objectives:

• Import songs and playlists from your music library
• Easy to use UI
• Play your songs and playlists
• Extract and manipulate the 4 core stems from your audio files as they play
• Discord rich presence to show off your music to your friends
• Save your stems as audio files
• If imported from your music library, view lyrics and metadata just as you would in your old music player
• Minimal memory usage
• Customizable themes
• Additional Efects like nightcore
• Easy to use equalizer
• Preprocessing service available on both CPU and GPU
• Ability to change the pre-trained model used for separation
• Small save size, comparable to the size of the inputted audio
• Docker image (WIP)
• Stable on Windows, Linux and MacOS (WIP)
• Proper installer/updater (Not a priority)

Installation

As of version 3.1.0, MISST is only available on windows with guaranteed compatibility.

Until a later release :

• if you are not on a windows device please refer to Manual Installation.
• if you are using conda on MacOs device please refer to Manual Installation - MacOS
• Otherwise, refer to the latest Release

Manual Installation

These instructions are for those installing MISST v3.1.0 manually only.

1. Download & install Python 3.9 or higher (but no lower than 3.9.0) here
   • Note: Ensure the "Add Python to PATH" box is checked
2. Download the Source code here
3. Open the command prompt from the MISST directory and run the following commands, separately -

$ python3 -m venv ./venv
$ pip install -r requirements.txt
$ python3 MISSTapp.py

• Note: Install requirements-minimal.txt if you don't intend to accelerate preprocessing with your GPU.

From here you should be able to open and run the MISSTapp.py file

• CUDA

  • CUDA must be installed and configured for the application to process any track with GPU acceleration. You will need to look up instruction on how to configure it on your operating system. Click here for nvidia's installation guide.

Benchmark

The audio processing performance was evaluated using an NVIDIA GeForce RTX 2070 SUPER with 8GB VRAM and an AMD Ryzen 3700X 8-Core Processor on the htdemucs pretrained model. This test aimed to compare the processing time of audio on a CPU versus a GPU.

Here are the results of the test:

Source	Source Length	CPU	GPU	Model
Frank Ocean - Ivy	4m 09.00s	2m 22.16s	0m 28.04s	htdemucs

Demo

MISST.mp4

Open in YouTube

License

The MISST code is GPL-licensed.

• Please Note: For all third-party application developers who wish to use MISST or its code, please honor the GPL license by providing credit to MISST and its developer.

Issue Reporting

Please be as detailed as possible when posting a new issue.

If possible, check the "MISST.log" file in your install directory for detailed error information that can be provided to me.

Contributing

• For anyone interested in the ongoing development of MISST, please send us a pull request, and I will review it.
• This project is 100% open-source and free for anyone to use and modify as they wish.
• I only maintain the development for MISST and the models provided.

More documentation to come...

misst's Issues

How to use Finetuned model "htdemucs_ft"

There exists the finetuned model 'htdemucs_ft', does yours have this feature? If not, how do I go about getting it on MISST?

Sound keeps going high and low

After I process a track and play it, it is very quiet and then gets VERY loud randomly through the whole track. Sounds the same after export. Tried multiple models. All do the same.

I used the exe to set up and did not download cuda.

Running and crash...

running in the middle and crashed and hanged.

MISST.log

MISST's Current State

In MISST's current state there are several issues that continue to make the app unusable to most and i'd like to enumerate them in this thread; As it stands MISST needs the following issues remedied to be a viable tool:

Server Issues:

• Very little uptime.
• Server flagged as virus or phishing website on some devices.
• Server is not scalable and in its current state can only handle one request at a time.

Processing Issues:

• No way to preprocess locally
  • I've tried to tackle this before but to preprocess locally the client would need a powerful GPU along with CUDA installed for any preprocessing to be done in sub-minute times.
• Only from-file, Spotify and Soundcloud imports are supported.

App Issues:

• Only compatible with windows devices.

Id like to invite anyone who sees this to contribute! if you have an idea feel free to make a PR, and if you have any questions DM me on discord or @ me in this thread.

Note: I will keep this thread updated with all pertinent information.

Server Down

I am aware the server has been down as of late and consequently it has been impossible to import new audios, when the V2.0.4 and some security updates are in place the server will be up. I will keep this issue updated with server info.

M1 MAC - Pyaudio problems - Could not import the PyAudio C module 'pyaudio._portaudio'.

Hello!

I have an M1MAX Macbook Pro with everything installed as a prerequisite.

I came across this and wanted to give it a try. The installation itself was problematic but i was able to fix the dependency issues with torchaudio and so.

Finally the app is built but i receive this after the python MISSTapp.py command:

(MISSTvenv) [~/dvlpmnt/MISST/MISST] python MISSTapp.py Could not import the PyAudio C module 'pyaudio._portaudio'. Traceback (most recent call last): File "/Users/erekhron/dvlpmnt/MISST/MISST/MISSTapp.py", line 25, in <module> from MISSThelpers import MISSTconsole, MISSThelpers File "/Users/erekhron/dvlpmnt/MISST/MISST/MISSThelpers.py", line 19, in <module> from MISSTplayer import MISSTplayer File "/Users/erekhron/dvlpmnt/MISST/MISST/MISSTplayer.py", line 8, in <module> import pyaudio File "/Users/erekhron/miniconda3/envs/MISSTvenv/lib/python3.12/site-packages/pyaudio/__init__.py", line 111, in <module> import pyaudio._portaudio as pa ImportError: dlopen(/Users/erekhron/miniconda3/envs/MISSTvenv/lib/python3.12/site-packages/pyaudio/_portaudio.cpython-312-darwin.so, 0x0002): symbol not found in flat namespace (_PaMacCore_SetupChannelMap) (MISSTvenv) [~/dvlpmnt/MISST/MISST]

After this i spent 3 hours to try to fix it but none of the google searched helped, i reinstalled stuff, reinstalled my whole python env etc.

Anything I could do? :(

Server offline

Hello, I went to test this tool, however, after installing it, when trying to import an audio, I get a message that the server is offline, how to fix it? thanks.

.

build on linux how to

hello
can you help me to build it under linux