Audio Fingerprinting

Поделиться
HTML-код
  • Опубликовано: 11 янв 2025

Комментарии • 7

  • @thedjsente5858
    @thedjsente5858 3 года назад +4

    Thanks for this great presentation and clear explanation. Would it be possible to get link to slides?

  • @JmanNo42
    @JmanNo42 4 года назад

    Isn't it possible to train an AI to recognise different instruments, fist starting with the insturments soloing finding out the hertz making notations sheets?
    I think there are projects that do transcribe music into midinotation and then it become pretty easy to check copyrights ?
    But that is maybe not what the project is about.

    • @firSound
      @firSound 4 года назад

      The term for what you are talking about is academically typically referred to as "Audio Source Separation".
      See: www.goodreads.com/book/show/36928190-audio-source-separation
      This is currently the best/most advanced commercial software for doing such analysis:
      ruclips.net/video/aPsxWSGEM0c/видео.html
      Then after the individual instruments are separated out into their respective stem track each,
      individual instruments can be converted into midi via Melodyne with the greatest precision.
      Melodyne can translate/transcribe both monophonic and/or polyphonic audio into Midi.
      ruclips.net/video/cvBYZPlitWE/видео.html
      Pretty nifty stuff to see in action if you haven't before, cheers.

  • @myolimpiada5037
    @myolimpiada5037 4 года назад

    wow!

  • @Maniclout
    @Maniclout 3 года назад +1

    The problem with storing hashes is that you can't turn it back into the original data.

    • @nixietubes
      @nixietubes Год назад +4

      that's the point, hashes/fingerprints are much much much lighter weight to search and store, it doesn't mean you can't also store the audio separately

    • @asandax6
      @asandax6 4 месяца назад

      That's the whole point of hashes. They are like titles for books, music, video etc so you can find it easier.