from Hacker News

Show HN: Polymath: Convert any music-library into a sample-library with ML

by samim on 2/11/23, 10:13 PM with 5 comments

Polymath is a open-source tool that converts any music-library into a sample-library with machine learning. It separates songs into stems, quantizes to same BPM, detects key and much more. A game-changing workflow for music producers & DJ

  • by westurner on 2/13/23, 11:27 PM

    Other cool #aiart things:

    - "MusicLM: Generating music from text" https://news.ycombinator.com/item?id=34543748

    - /? awesome Generative AI site:GitHub.com https://www.google.com/search?q=awesome+generative+ai+site%3...

    - #GenerativeArt #GenerativeMusic

    - BespokeSynth DAW: https://github.com/BespokeSynth/BespokeSynth :

    > [...] live-patchable environment, so you can build while the music is playing; VST, VST3, LV2 hosting; Python livecoding; MIDI, [...]

  • by westurner on 2/13/23, 11:05 PM

    > Using Polymath's search capability to discover related tracks, it is a breeze to create a polished, hour-long mash-up DJ set

    https://github.com/samim23/polymath#how-does-it-work :

    > How does it work?

    > - Music Source Separation is performed with the [facebook/demucs] neural network

    > - Music Structure Segmentation/Labeling is performed with the [wayne391/sf_segmenter] neural network

    > - Music Pitch Tracking and Key Detection are performed with [marl/Crepe] neural network

    > - Music Quantization and Alignment are performed with [bmcfee/pyrubberband]

    > - Music Info retrieval and processing is performed with [librosa/librosa]

  • by an_aparallel on 2/12/23, 10:45 AM

    Bloody hell...seriously...the rate at which stuff comes out is just baffling. samim - do you have examples of this in use? I've kinda dreamed of something like this for some time!