CHANGELOG.md

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog.

v3.21.0 - 2024-08-16

Changed

• Update GPU Dockerfile to CUDA 11.8 and PyTorch to 2.4.0
• Update dependencies

v3.20.2 - 2024-06-28

Added

• Add missing dependency for building on AArch64

Changed

• Fix iOS support (#1145)
• Update dependencies

v3.20.1 - 2024-02-01

Changed

• Fix regression where first-time Spleeter model download fails due to httpx not following redirects (#949) - thanks @microtherion for the fix!
• Add cleanup logic to remove leftover model files from failed download attempts
• Update dependencies

v3.20.0 - 2024-01-13

Added

• Add Docker support for AArch64 systems (including Apple Silicon)

Changed

• Update Docker Python version to 3.9
• Update Docker CUDA image version to 11.6.1
• Update dependencies

v3.19.0 - 2023-09-24

Added

• Add 2 environment variables for configuring yt-dlp:
  • YOUTUBEDL_SOURCE_ADDR - Client-side IP address for yt-dlp to bind to. If you are facing 403 Forbidden errors, try setting this to 0.0.0.0 to force all connections through IPv4
  • YOUTUBEDL_VERBOSE - Set to 1 to enable verbose logging for yt-dlp
• Add download button on music player

Changed

• Update dependencies

v3.18.0 - 2023-07-01

Added

• Add AWS_S3_REGION_NAME and AWS_S3_SIGNATURE_VERSION variables for S3 integration
• Add more descriptive error logging

Changed

• Allow multiple CERTBOT_DOMAINs and hardcode cert-name to 'spleeter-web' (#533)
• Fix bug where separation would fail immediately (#672)
• Update GPU Docker base image to cuda:11.2.2 from cuda:11.2.0 due to deprecation
• Update dependencies

v3.17.0 - 2023-01-11

Added

• Support Spleeter's 5-stem model which separates piano in addition to other accompaniment

Changed

• Update separator form UI to use buttons instead of dropdown menus
• Improve form field validation
• Bring back 'iterations' param for X-UMX model
• Update dependencies
• Fix Dev Docker Compose config
• Fix bug where deleting dynamic mix may cause blank screen

Removed

• Remove frontend upload size check and rely on Nginx and backend instead

v3.16.0 - 2022-12-26

Added

• Support lossless output formats (WAV, FLAC)
• Add UPLOAD_FILE_SIZE_LIMIT (in MB) and YOUTUBE_LENGTH_LIMIT (in minutes) environment variables to customize upload limits

v3.15.0 - 2022-12-11

Added

• Update Demucs to v4, which features new Hybrid Transformer models with average SDR of 9.0
  • More info
• Add ability to configure segment sizes for Demucs models
  • If Demucs separation fails for you due insufficient GPU memory, try setting the environment variable DEMUCS_SEGMENT_SIZE to a lower value like 10 (default is 40, which requires a around 7 GB of memory). You may have to experiment a bit to find the appropriate value. See this for more info.
  • Also try setting PYTORCH_NO_CUDA_MEMORY_CACHING=1 to disable caching

Changed

• Update Docker Compose definitions from version 3.4 to Compose Specification
  • You may have to update your Docker Compose version to the newer version

v3.14.0 - 2022-09-08

Changed

• Raise minimum Python version to 3.8 and NodeJS version to 16
• Update Docker images to use Python 3.8 and NodeJS 16 (bullseye)
• Update Spleeter, Demucs, Tensorflow, and other dependencies

Removed

• Remove some unused NPM dependencies

v3.13.0 - 2022-06-23

Changed

• Update dependencies
• Docker: start Django server using 'api' instead of 0.0.0.0 as bind address

Added

• Support for RTX GPUs - Thanks @Ma5onic! (#142)
• HTTPS support using docker-nginx-certbot.
  • To use HTTPS, set APP_HOST to your domain name and CERTBOT_EMAIL to your email in .env and include -f docker-compose.https.yml in your docker-compose up command.
• Add ability to export Dynamic Mixes based on each component's volume levels.
  • To enable this feature, set ENABLE_CROSS_ORIGIN_HEADERS=1 in .env and either access Spleeter Web at localhost or enable HTTPS on your domain (see above). This is because it uses SharedArrayBuffers which require cross-origin isolation.
    • If using an external storage provider, you'll need to set the Cross-Origin-Resource-Policy response headers to cross-origin. See this for more details.
  • All the exporting is done in-browser using FFmpeg.WASM.
  • Sidenote: Spleeter Web uses a forked version with some cherry-picked changes from the main FFmpeg repo pertaining to the amix filter.

v3.12.0 - 2022-02-21

Changed

• Update X-UMX and D3Net models to latest version
• Fix bug where status icon overlay sometimes disappears
• Fix bug where adding basic auth in front of nginx breaks the app (credit to @jtagcat)
• Update dependencies

Added

• Add API_HOST environment variable to control hostname of API server for nginx (credit to @jtagcat)
• Show datetime when fetch task or separation task finished on status icon hover
• Model file integrity checks for X-UMX and D3Net models
• Add support for accelerated CPU separation for D3Net using OpenVINO
  • To enable, set D3NET_OPENVINO environment variable to 1 and D3NET_OPENVINO_THREADS to number of CPU threads to use

Removed

• Prune unused pip dependencies

v3.11.0 - 2021-12-17

Changed

• Update Demucs to v3, which uses hybrid separation approach resulting in much improved performance
  • Official Demucs changelog
• Fix bug where non-Spleeter models always generated 128kbps MP3 files. Oops!
• Update "Getting Started" section in README to use prod configuration
• Update some dependencies

Removed

• Tasnet models
• Ability to separate using older Demucs models

v3.10.0 - 2021-11-01

Changed

• D3Net - Fix "Unable to open file (file signature not found)" error
• D3Net - Fix separation when using local file storage
• Make DEFAULT_FILE_STORAGE an environment variable. It can be set to FILE, AWS, or AZURE. More information in README.md.
  • Set the default to FILE in all cases
• Apply workaround for DNS issue on Docker for Windows

v3.9.0 - 2021-10-09

Changed

• Update Spleeter to 2.3.0 (which uses Tensorflow 2.5.0)
• Update GPU Dockerfile to use CUDA 11.2.0/cuDNN8 and latest nnabla package
• Switch from youtube_dl (inactive project) to yt_dlp which fixes slow YouTube download speeds
• Fix issue with using Docker on Windows where YouTube videos would not download
• Update dependencies

v3.8.0 - 2021-06-25

Added

• Docker: Add environment variables DEV_WEBSERVER_PORT and NGINX_PORT to configure which ports to use on host machine for webserver

Removed

• Docker: Unexpose Redis port 6379 on host machine

v3.7.0 - 2021-06-21

Added

• Support for D3Net model! Big thanks to Sony AI Research for making it open-source (paper).

Changed

• Fix GPU Docker image so that GPU-accelerated separation works properly with Spleeter
• Update youtube-dl
• Fix YouTube metadata parsing on certain types of videos

v3.6.1 - 2021-06-14

Changed

• Add back Demucs Light variant mapping
• Update more dependencies (security fixes)

v3.6.0 - 2021-06-13

Changed

• Update Demucs to v2, which offers higher quality separations and smaller models
• Allow all hosts when running in dev mode
• Update dependencies

v3.5.2 - 2021-04-16

Changed

• Update dependencies

v3.5.1 - 2021-02-15

Changed

• Suppress discovery_cache warnings related to YouTube data API calls
• Update dependencies

v3.5.0 - 2021-02-07

Added

• Ability to download individual parts of a dynamic mix

Changed

• Action buttons appear greyed out after pressed while waiting for backend API response

v3.4.0 - 2021-02-06

Added

• Support additional audio formats
  • Full list: .aac, .aif, .aifc, .aiff, .flac, .m4a, .mogg, .mp3, .oga, .ogg, .opus, .wav, .weba, .webm
• Ability to edit source track artist and title information
  • In the song table, just click on the cell to edit its value. Then press enter or click elsewhere to save.

Changed

• Improve YouTube link parsing logic
• Increase max file size upload limit to 100 MB
  • You can override this by changing UPLOAD_FILE_SIZE_LIMIT in settings.py/settings_docker.py and MAX_FILE_BYTES in Constants.tsx
• Increase max YouTube video length to 30 minutes
  • You can override this by changing YOUTUBE_LENGTH_LIMIT in settings.py/settings_docker.py
• Update Node dependencies

v3.3.1 - 2021-01-30

Changed

• Fix bug where Spleeter static mixes would fail

v3.3.0 - 2021-01-18

Added

• Support X-UMX CPU separation

Changed

• Make dynamic mix player timestamp more consistent

v3.2.0 - 2021-01-17

Added

• Solo track functionality to dynamic mixer
• Keyboard controls for dynamic mixer
  • Play/pause: spacebar
  • Mute/unmute: 1/2/3/4
  • Solo/unsolo: Q/W/E/R (hold either Ctrl/Cmd/Shift to solo/unsolo multiple parts)

Changed

• Pressing space bar on home page should play/pause current track without scrolling to bottom
• Spleeter: Update to v2.1.2
• Spleeter: use STFTBackend.LIBROSA for CPU separation and STFTBackend.TENSORFLOW for GPU separation

v3.1.1 - 2021-01-14

Changed

• Update react-music-player with bug fixes

v3.1.0 - 2021-01-13

Added

• Environment variables to support custom domains such as for CDNs

Changed

• Use "title - artist" format everywhere
• Update react-music-player to use non-linear volume slider and gradual fade-ins/fade-outs
• Update dependencies

v3.0.0 - 2021-01-08

Added

• GPU-enabled Docker images (CUDA 10, CUDNN 7) which can accelerate separation process
  • Use docker-compose.gpu.yml and docker-compose.build.gpu.yml instead of docker-compose.yml and docker-compose.build.yml for GPU images
  • See updated instructions for more info
  • Tested on NC6 Promo Azure VM
• Support for Sony AI's CrossNet-Open-Unmix (X-UMX) source separation model (NNabla implementation). At the moment it only works with GPU, not CPU. #52 tracks this issue.
  • User configurable parameters: number of iterations, softmask, softmask alpha
• Custom bitrates (MP3 CBR) for static and dynamic mixes at: 192 kbps, 256 kbps, and 320 kbps

Changed

• Improved dynamic mix processing times by parallelizing the MP3 export process
• Reduced memory consumption of Demucs/Tasnet models by using splitting
• Update dependencies

Removed

• Overwrite option for static mixes (you can always delete specific mixes)

v2.0.1 - 2020-12-30

Changed

• Indent mix table to make it easier to see
• Update Demucs dependency

v2.0.0 - 2020-12-29

Added

• Support for Facebook Research's Demucs and Tasnet source separation models, including "light" and "extra" variants
  • Demucs and Tasnet models have a "random split" parameter that can help improve separation quality
• New labels to indicate the model and parameters used to generate a mix
• New refresh button that also indicates the time remaining until next auto-refresh
• Add brief fade-in and fade-out to dynamic mix player
• Show separator badge in music player

Changed

• Dynamic mixes now appear as part of the mix table
• Use more descriptive and cleaner file naming scheme for generated files
• Minor touch-ups to overall interface
• Fix bug where sort order of mix table gets reset after a few seconds
• Fix "'ContentDisposition' is an invalid key" error when writing to Azure storage
• Fix bug where music player timestamp would show "00:60" instead of "01:00"
• Fix blank status badge

v1.2.0 - 2020-12-22

Added

• New configurable settings for using AWS S3 as backend for serving media files
• A favicon

Changed

• Files uploaded to Azure Storage have Content-Disposition always set to attachment
• Fix issue with Dynamic Mixes where seeking occurs after long delay

Removed

• psycopg2 Python dependency

v1.1.0 - 2020-12-20

Breaking Changes for Docker

• Change database backend from PostgreSQL to SQLite
  • If you are updating from a previous version, please backup your track list as the data in the DB will not carry over after updating! Your media files will not be impacted.

Added

• On dynamic mix pages, the tab title shows the current track information

Changed

• Dynamic mix files are now saved as: artist-title-part.mp3 instead of part.mp3
• Correct service name in docker-compose.prod.selfhost.yml
• Use separate Celery queues for fast (YouTube imports) and slow (source separation) tasks
  • For Docker, the fast and slow Celery workers run in separate containers
• Update dependencies

v1.0.0 - 2020-12-19

Added

• Ability to delete individual static mixes
• Ability to delete source tracks and mixes while task is still in-progress
• Ability to cancel in-progress dynamic mix task
• New status icon column in the Track List table
• New requirement for Redis
• Make AZURE_CONTAINER a configurable environment variable

Changed

• Fix bug where dynamic mix tracks may never finish loading
• Dynamic mixes open in new tab
• Change file deletion logic to also delete empty parent directories
• Make the "external link" icon a button in the YouTube search result list
• Switch from Huey to Celery for the task queue
  • Celery allows terminating in-progress tasks
• Increase YOUTUBE_LENGTH_LIMIT to 20
• Decrease API polling frequency to 5 seconds
• Update Spleeter to 2.0.2 (Python 3.8 now supported)
• Update Python and npm dependencies
• Update Docker images

Removed

• Periodic cleanup task

Pre-release - before 2020-12

Undocumented