How it works

SoundBud does two jobs. Decode turns a song into a chart you can play. Polish takes your own recording to studio sound. Video editors focus on the picture, SoundBud focuses on the sound.

Decode

Turn a song into a chart you can play.

  1. Bring in the song.

    Record it live, pick a file, or send it from any app with the iOS share sheet. Video works too: SoundBud reads the audio out of it.

  2. Get the chart.

    In seconds you get the key, BPM, time signature and time-stamped chords, laid out as bars like a lead sheet. Any value looks wrong? Tap it and fix it, the grid recalculates instantly.

    Harmonic analysis in SoundBud: chords colour-coded as diatonic, borrowed or secondary dominant with Roman numerals underneath, above a bar grid reading E major, 3/4, 120 BPM.

  3. Play it.

    Slow it down, loop a section, transpose it, hear the chords on a synth, run the metronome. Then pick an AI mode: Theory, Solo Guide, Reharmonize or Explain Simply.

    A finished SoundBud session: detected key E major, 120 BPM and 3/4 meter, genre and instrument tags, a spectrum visualiser, playback controls and the Practice tools bar.

Polish

Take your own audio to studio sound.

  1. Drop in your audio.

    A recording, a podcast, a voice memo, or a whole video. Trim it losslessly first if you want.

  2. Shape it.

    Four tabs: Voice-Over (de-esser, breath, clarity, room tone), Ducking (music under speech), EQ and Dynamics with AI tips that measure your actual audio, and Output.

    The Voice-Over tab in SoundBud Polish, with De-Esser, Breath Removal and Clarity Boost, each a toggle with an amount slider.
    The EQ and Dynamics tab in SoundBud Polish, showing AI Tips, Voice, Music and Field presets, and a 6-band equaliser from Low Cut to High Cut.

  3. Export it.

    Pick your platform target (Spotify, YouTube, TikTok, Podcast) and SoundBud sets the right loudness. Export WAV or MP3, or get your video back with the same picture and better sound.

    The Output tab in SoundBud Polish: platform presets with their loudness targets, and a LUFS comparison moving the track from its current level to the selected target.

Under the hood

  • Chords are detected by an ensemble of neural networks across 301 chord classes, covering a wide vocabulary well beyond simple triads: sevenths, ninths, sus and slash chords.
  • Tempo and beat positions come from a separate rhythm model.
  • Genre and instrument detection use a separate classification model.
  • The key is not guessed by AI. It is derived from the detected chords on your device, using a classic music theory method (Krumhansl profile matching).
  • Polish runs entirely on your device. No internet needed.
  • MIDI export runs on your device.
  • Song recognition uses the AudD music recognition service.
Download on App Store