1.1a1

1.1a1

For Apple Silicon users: please use Demucs-GUI_1.1a1_macOS_x86_64_Rosetta2.dmg due to a bug mentioned in README. It can reach 25% performance compared with native ARM64 software if you separate an audio with CPU, but MPS acceleration is available and will not affect the accelerated speed.

Download site FossHUB is also available.

If the program can't start up on macOS, please follow the instructions here.

Updates

1. Use the new tabbed UI
2. Add new option: separate once added to the queue
3. Save settings history for separate-once-added, file format, save location and clip mode

Fixes

1. Will show download progress if a remote model is not downloaded

In previous versions, Demucs-GUI will not change the default torch hub cache dir. However, from this version, all new models will be downloaded to Demucs-GUI.config.dir/pretrained/checkpoints, without copying the old ones there. So you need to copy it manually if you still need those models.

2. Fix auto check updates doesn't work
3. Fix model info can't be shown completely sometimes

Full Changelog: 1.0.2.1...1.1a1

1.0.2.1

1.0.2.1

Emergency fix for 1.0.2

For Apple Silicon users: please use Demucs-GUI_1.0.2.1_macOS_x86_64_Rosetta2.dmg due to a bug mentioned in README. It can reach 25% performance compared with native ARM64 software if you separate an audio with CPU, but MPS acceleration is available and will not affect the accelerated speed.

Download site FossHUB is also available.

If the program can't start up on macOS, please follow the instructions here.

Fixes

1. Fix an issue about reading audio with FFMpeg

Known issues

1. May causes waiting for long time when loading a remote model for the first time

Full Changelog: 1.0.2...1.0.2.1

1.0.2

1.0.2

For Apple Silicon users: please use Demucs-GUI_1.0.2_macOS_x86_64_Rosetta2.dmg due to a bug mentioned in README. It can reach 25% performance compared with native ARM64 software if you separate an audio with CPU, but MPS acceleration is available and will not affect the accelerated speed.

Download site FossHUB is also available.

If the program can't start up on macOS, please follow the instructions here.

Updates

1. Show submodels in bag of models before loading it

Fixes

1. Fix an issue about reading audio

Known issues

1. [NEW] [ADDED] Will get stuck when reading with FFMpeg
2. May causes waiting for long time when loading a remote model for the first time

Full Changelog: 1.0.1...1.0.2

1.0.1

1.0.1

For Apple Silicon users: please use Demucs-GUI_1.0.1_macOS_x86_64_Rosetta2.dmg due to a bug mentioned in README. It can reach 25% performance compared with native ARM64 software if you separate an audio with CPU, but MPS acceleration is available and will not affect the accelerated speed.

Download site FossHUB is also available.

If the program can't start up on macOS, please follow the instructions here.

PyTorch has been downgraded to 2.0.1 for better compatibility.

Updates

1. Automatically check updates

Fixes

1. Fix an issue that output is always rescaled to 99.9% when clip mode is set to "rescale"
2. Fix an issue that window will stop to function when opening log file on non-Windows
3. Fix an issue related with detecting FFMpeg on Windows
4. Optimize ETA algorithm

Known issues

1. May causes waiting for long time when loading a remote model for the first time

Full Changelog: 1.0...1.0.1

1.0

1.0

For Apple Silicon users: please use Demucs-GUI_1.0_macOS_x86_64_Rosetta2.dmg due to a bug mentioned in README. It can reach 25% performance compared with native ARM64 software if you separate an audio with CPU, but MPS acceleration is available and will not affect the accelerated speed.

Download site FossHUB is also available.

If the program can't start up on macOS, please follow the instructions here.

From this release, FFMpeg will be included in the package, which makes the binaries larger.

Updates

1. Read with FFMpeg (which allows separating more file formats, even extracting audio track from a video and separate it)
2. Save files with multithread to avoid waiting between tracks
3. Show ETA
4. Ask before quitting if separating else exits (Yes, this is a Python conditional expressions)
5. Logging more separating parameters
6. Allow changing window style using settings.json
7. Allow PyQt6 as backend (though packed binaries will not use it. You can use it by modifying shared.py)
8. Add demucs_unittest model
9. Add menu bar

Fixes

1. Use absolute path for font
2. Fixed init error of mps
3. Fix sometimes progress goes back
4. Force separation queue using Fusion style on macOS to show progress bar

Known issues

1. May causes waiting for long time when loading a remote model for the first time
2. [ADDED] [FIXED] Could not use remote models due to a mistake packing the application on Windows
3. [ADDED] Can't start FFmpeg sometimes. If you see FFMpeg is not available, please restart the application.
4. [ADDED] Will always rescale to 99.9% when clip mode is set to "rescale"

New Contributors

• @HallowedDust5 made their first contribution in #20
• @kyab made their first contribution in #32

Scans (by VirusTotal)

Only the main executable (no libraries and other support files) is submitted to VirusTotal.

1. Demucs-GUI_1.0_cpu
2. Demucs-GUI_1.0_cuda
3. Demucs-GUI_1.0_macOS_x86_64_Rosetta2

Full Changelog: 1.0a1...1.0

1.0a1

1.0a1

For Apple Silicon users: please use Demucs-GUI_1.0a1_macOS_x86_64_Rosetta2.dmg temporarily. It may slower than pure ARM64 software if you separate an audio with CPU, but it can use MPS acceleration.

If the program can't start up on macOS, please follow the instructions here.

Updates

1. Rewritten GUI with PySide6 (Qt)
2. Listing model support
3. Getting model details
4. Multi-channel support - separate each channel one by one and combine them back
5. Separation queue support - automatically separate each song one by one
6. Drag & Drop support - simply drag folders or files into the queue window and will be added to queue
7. Progress bar
8. More save options - flac/wav, clamp/rescale, int16/int24/float32

Known issues

1. Waiting for documentation! Everybody is welcome to contribute! (#23)
2. May causes waiting for long time when loading a remote model for the first time
3. No ffmpeg support though I've added ffmpeg detection
4. [ADDED] [FIXED] Packed binaries can't use remote models
5. [ADDED] macOS can’t start up due to relative path

Full Changelog: 0.1...1.0a1

0.1

0.1

Release date: Jun. 22, 2022

About the downloads

We provide binaries for Windows 64-bit and macOS 64-bit. Apple silicon binaries is not currently available.
For Windows, we also packed two files depending on whether CUDA is enabled or not, so you can choose CUDA version if you have a NVIDIA graphic card with compute capability >= 3.1. It is about at least 20 times faster than CPU when CUDA is enabled.

Both binaries are packed with PyInstaller 5.1. Binaries for Windows is built under Windows 11, Python 3.10. Binaries for macOS is built under macOS 10.15, Python3.9.

We recommend downloading from FossHub to speed up, but please read this release note first.

File list

1. Demucs-GUI_0.1_CUDA_Release_Win_x64.7z

This is the pack with CUDA enabled and mdx_extra_q model. You can simply run Demucs-GUI.exe.

Size: 1,371,584,957 Bytes (1,308 MiB)
SHA1: F6FF3CC775F94A71AA41DF6ADE16CCBA764EFB38
Decompress size: 4,517,900,628 (4,308 MiB)

2. Demucs-GUI_0.1_CPU_Release_Win_x64.7z

This is the minimal pack with only CPU and mdx_extra_q model. You can simply run Demucs-GUI.exe.

Size: 241,511,386 Bytes (230 MiB)
SHA1: FD6DB045AE37ABEBCD0E309ABADAEB02EA415FB8
Decompress size: 1,052,168,109 (1,003 MiB)

3. Demucs-GUI_0.1_Release_macOS_x64.dmg

This is the pack for macOS x64. Apple Silicon is not supported.

Size: 424,908,870 Bytes (425MB)
SHA1: EC9E2F7D95C2AE20E4D54B0BAD470015E68869EB
Install size: About 1.14GB

Why I can't run the program?

Because of the limitation of Apple, Demucs-GUI need some extra configuration to work properly.

First, we should allow running apps from all of sources. Execute following command in your Terminal (if you do not know where your Terminal.app is, please search your dashboard):

sudo spctl --master-disable

You may need to input your password.

Then, we need to bypass the notarization (replace the path below to where your Demucs-GUI.app is if you did not install to the default location):

sudo xattr -rd com.apple.quarantine /Applications/Demucs-GUI.app

Alternative download sites

FossHub

https://www.fosshub.com/Demucs-GUI.html

Support us

The reason to the problem above is that this application is not certificated. Since certification requires becoming Apple Developper, which is not free, we currently cannot certificate it. Besides, we don't have a M1 MacBook, so those binaries are not available. If you want to help with that, please open an issue to tell us and we will tell you what to do to generate the binaries.

Updates

1. Solved mono audio reading issue
2. Now you can type in the numbers of splits and overlap.
3. Windows now can read ogg and mp3 files.
4. Increase log level

Known issues

1. FFMpeg is still not available
2. On macOS, reading of mp3 files is not available. We are waiting the 0.11.0 release of python-soundfile.

Full Changelog: 0.1a2...0.1

0.1a2

0.1a2

Release date: May. 10, 2022

This is a beta version so binaries will not be packed.
Please follow the guide under README.md to run.

Maybe you feel the code style is not readable, we will fix that when releasing the first formal release.

Updates

1. Added logging so it will be friendlier to debugging
2. Removed requirement for pynvml and torchaudio
3. Thanks @hanton2222 , so logging for macOS could be initialized (#8)

Known issues

1. Since the problem of shifts (#6), shifts is temporarily removed
2. FFMpeg is still not available

Full Changelog: 0.1a1...0.1a2

0.1a1 initial release

0.1a1

Release date: Mar. 31, 2022

This is a beta version so binaries will not be packed.
Please follow the guide under README.md to run.

Maybe you feel the code style is not readable, we will fix that when releasing the first formal release.

Features

1. Graphic User Interface available
2. Updates system usage info every 0.2 second
3. Reads your hardware information and gives choices
4. Redeces least VRAM requirement to 3GiB
5. Provides recommended value for split based on your hardware

Known issues

1. Though the welcome page tells whether FFMpeg is installed, FFMpeg would not be used.
2. I forgot to add psutil>=5.7.0 into requirements.txt and requirements_cuda.txt, please add it manually before installing dependencies.

Full Changelog: https://github.com/CarlGao4/Demucs-Gui/commits/0.1a1