Sign in

v1

latestOpenAPI 3.0.02026-07-26355125.0 KB

Generate MIDI from Audio

Convert separated audio tracks into MIDI format with detailed note information for each instrument.

Usage Guide

  • Convert separated audio tracks into structured MIDI data containing pitch, timing, and velocity information
  • Requires a completed vocal separation task ID (from the Vocal Removal API)
  • Generates MIDI note data for multiple detected instruments including drums, bass, guitar, keyboards, and more
  • Ideal for music transcription, notation, remixing, or educational analysis
  • Best results on clean, well-separated audio tracks with clear instrument parts

Prerequisites

  • You must first use the Vocal & Instrument Stem Separation API with type: split_stem to separate your audio before generating MIDI

Parameter Details

  • taskId identifies a completed vocal separation task
  • callBackUrl receives MIDI generation completion notifications
  • audioId (optional) specifies which separated audio track to generate MIDI from. Obtain this ID from the originData array in the Get Vocal Separation Details response. If not provided, MIDI will be generated from all separated tracks.

Developer Notes

  • MIDI generation typically takes 30-90 seconds depending on audio length and complexity
  • The callback will contain detailed note data for each detected instrument
  • Each note includes: pitch (MIDI note number), start (seconds), end (seconds), velocity (0-1)
  • Not all instruments may be detected - depends on audio content
  • Generated MIDI data is retained for 14 days
post/api/v1/midi/generate

Request body

taskIdstring required

Task ID from a completed vocal separation. This should be the taskId returned from the Vocal & Instrument Stem Separation endpoint.

callBackUrlstring uri required

The URL to receive MIDI generation task completion updates. Required for all MIDI generation requests.

  • System will POST task status and MIDI note data to this URL when generation completes
  • Callback includes detailed note information for each detected instrument with pitch, timing, and velocity
  • Your callback endpoint should accept POST requests with JSON payload containing MIDI data
  • For detailed callback format and implementation guide, see MIDI Generation Callbacks
  • Alternatively, use the Get MIDI Generation Details endpoint to poll task status
audioIdstring

Optional. Specifies which separated audio track to generate MIDI from. This audioId can be obtained from the originData array in the Get Vocal Separation Details endpoint response. Each item in originData contains an id field that can be used here. If not provided, MIDI will be generated from all separated tracks.

Example request

{
  "taskId": "5c79****be8e",
  "callBackUrl": "https://example.callback",
  "audioId": "8ca376e7-******-08aaf2c6dd27"
}

Response

MIDI generation task created successfully

codeinteger

Response status code

msgstring

Response message

Example response

{
  "code": 200,
  "msg": "success",
  "data": {
    "taskId": "5c79****be8e"
  }
}