Improved freemusicdemixer
AI
music source separation
SHOW HN2023
160points
40comments
#379of 3,109 in 2023
Top 13%of 2023 by points
How HN took it
DebatedReal support and real criticism
9 positive, 4 neutral or asking, 1 mixed, 7 critical, out of 21 comments about the project. Thread score +0.07.
Execution: 1 for, 6 againstIdea: 5 for, 0 againstPrivacy: 2 for, 0 againstPerformance: 0 for, 1 against
The thread, comment by comment
Each row is a top-level comment and its replies, in HN's order. Select a cell to open that comment on HN.
positive neutral or mixed critical about another commenter or something else the maker
What HN said
What people liked
Runs free and locally in the browser with no upload
From 3 comments
Man, this is fucking awesome. Thank you for building this! Not just free, but local, without installation, and generalized to common use-cases too? This is a standard of…
catapart on HN"Runs locally in your browser! Unlike similar products, it’s free to use and doesn’t store your data. All processing is done in your browser, and your files are never uploaded…
dylan604 on HNIt is very nice to be able to compute this in our browsers without limits or registration. The results are not always perfect but already allows having good time testing remixes…
tkp on HN
Useful for remixes, karaoke and practice
From 3 comments
Man, this is fucking awesome. Thank you for building this! Not just free, but local, without installation, and generalized to common use-cases too? This is a standard of…
catapart on HNIt is very nice to be able to compute this in our browsers without limits or registration. The results are not always perfect but already allows having good time testing remixes…
tkp on HNWhile this might seem to be intended for DJ remixes a demixer also lets you jam along to your favorite songs. For instance you can extract only the voice, bass, and drums and…
nre on HN
Objections
Stems show audible generative smearing compared to non-AI separation
From 1 comment
after finally getting some free time to play with this, i did. i've used other non-AI based programs to remove vocals. the in sound is totally different and obviously beyond…
dylan604 on HN
Fails or runs out of memory on longer tracks and some Chrome setups
From 4 comments
I frequently use a similar online demixer and am always worried that they will eventually remove the free option, so was excited to try this one out. When uploading couple…
jimmydroptables on HNJust tried this and it seems to get stuck at the following step (for 20 mins so far) [WASM/C++ 18:07:24] Getting waveforms from istft [WASM/C++ 18:07:25] Copying waveforms
phpnode on HNah out of memory, I'll try with a shorter track.
phpnode on HNI got same, but have 40900 and 32 Gig. On multiple files.
airbreather on HN
Slow or unusable on smartphones
From 1 comment
"Runs locally in your browser! Unlike similar products, it’s free to use and doesn’t store your data. All processing is done in your browser, and your files are never uploaded…
dylan604 on HN
Requests
Support for longer tracks and clearer error or length-limit messages
From 2 comments
I frequently use a similar online demixer and am always worried that they will eventually remove the free option, so was excited to try this one out. When uploading couple…
jimmydroptables on HNah out of memory, I'll try with a shorter track.
phpnode on HN
Demucs-quality model running in WASM
From 1 comment
My dream (and next major project) is to get Demucs (v4, Hybrid Transformer) running in WASM, similar to what I've done with Eigen/C++ and Open-Unmix.
sevagh on HN
VST plugin frontend for DAWs
From 1 comment
Amazing work Would love a VST frontend as well so I could drop a song file into a plugin and have it spit the stems out directly into my DAW
whoislewys_1 on HN
Better labels than Other for the stems
From 2 comments
Labeling the 4 parts Vocal | Melody | Bass | Drums, instead of Drums | Vocals | Bass | Other would go a long way to making it seem less programmer-y.
fragmede on HNNice idea! The name 'other' comes from the traditional datasets used to train these models, but I'm open to trying to incorporate a better name.
sevagh on HN
Questions
Is this the best quality possible or a reduced model for browsers?
From 1 comment
Is this the best possible demixing you can get? Or did you have to use a smaller / lower quality model to make it run in browsers?
Spiwux on HN
Can tools add accompaniment to a vocal track?
From 1 comment
Are there any tools that will add musical accompaniment to a vocal track? Basic strings/piano?
lh15 on HN
Alternatives named
Demucs
From 3 comments
For those interested, Facebook's Demucs page (https://github.com/facebookresearch/demucs) gives performance comparison for several models including open-unmix. See also:…
alok-g on HNBy my own little comparison in which I've given the same song to both, Demucs appears much better (while requiring an app and consuming far more CPU). Thanks for sharing! I had…
input_sh on HNdemucs works far better than anything else I've used, especially with more esoteric kinds of music in my experience. plus you can run it with GPU support as well!
adzm on HN
StemRoller
From 2 comments
For those interested, Facebook's Demucs page (https://github.com/facebookresearch/demucs) gives performance comparison for several models including open-unmix. See also:…
alok-g on HNOh my god, stemroller.app is 1.85 GB (weights only 350 mb) electron python pytorch ffmpeg and 10 MB webapp. This is state of modern app development. But it works, nonetheless.
diimdeep on HN
umx.cpp
From 1 comment
It didn't load weights for me, So I tried umx.cpp but it failed to compile on mac. I would recommend to polish umx.cpp and put converted weights on huggingface, make it as…
diimdeep on HN