Audio is the source
The recording contains the sound of the performance. Keep it available so you can compare it with the MIDI after conversion.
Melogen / MCP
For existing SSE/API key clients. This integration is not included in the public Melogen Music Intelligence ChatGPT plugin.
Manage legacy API keyLegacy SSE/API key integration
Three checks for an existing legacy integration.
Use the SSE endpoint and API key already configured for your legacy client. This conversion tool is separate from the public ChatGPT plugin and its OAuth connection.
Provide the audio file URL, not a local file path or a page that requires your browser login. The legacy service must be able to retrieve the file.
The example uses save_stems: false and quantize: false. Request stems only when your legacy service supports them; enable quantization when you want note timing aligned to a beat grid.
Use this path only if you already have a legacy SSE/API key integration; the public ChatGPT plugin uses a different product flow.
Keep your existing SSE endpoint and API key configuration. Do not use this configuration for the public ChatGPT plugin.
Ask your compatible AI client to call convert_audio_to_midi with an audio URL accessible to the legacy service.
Use the result returned through your existing client workflow, then review and edit the MIDI as needed.
Concept illustration — these drawings are not a conversion result.
The recording contains the sound of the performance. Keep it available so you can compare it with the MIDI after conversion.
Open the returned MIDI in your MIDI editor or DAW, choose an instrument, then listen against the source. Review note pitches, timing and lengths before using it in an arrangement.
Listen to a short passage at a time. Check overlapping notes, missed notes and timing around vocals or layered instruments. A returned MIDI file is a result to review, not a guarantee of an exact transcription.
Illustrative prompts for an already configured legacy client. Replace the URL with your own accessible audio file.
Use convert_audio_to_midi with my audio URL. Set save_stems and quantize to false. Return the MIDI result so I can review it.
A starting point for checking the transcription before deciding on further processing.
Convert my audio URL to MIDI with quantize set to true and save_stems set to false. I want to review the notes against a beat grid.
Quantization changes note timing. Compare the result with the original if the performance uses expressive timing.
Illustrative request and response only. The sample task ID and download URL below do not represent a real job.
Existing SSE/API key clients can use this legacy tool to convert an audio file URL to MIDI. It is separate from the public Melogen Music Intelligence ChatGPT plugin.
Audio supplied by URL: MP3, WAV, FLAC, OGG, M4A Output: MIDI
file_url (required) — a URL the legacy service can accesssave_stems — request separated instrument stems when supportedquantize — align MIDI notes to nearby beats when neededThis prompt is for an already configured legacy SSE/API key client, not the public ChatGPT plugin.
Example prompt:
Use convert_audio_to_midi to convert this accessible audio URL to MIDI. Keep save_stems and quantize disabled unless I request them.
// MCP Tool Call (handled automatically by your AI agent)
{
"tool": "convert_audio_to_midi",
"arguments": {
"file_url": "https://example.com/song.mp3",
"save_stems": false,
"quantize": false
}
}
// Response
{
"task_id": "xyz789",
"status": "completed",
"download_urls": {
"midi": "https://api.melogenai.com/files/xyz789.mid"
},
"message": "Audio converted to MIDI successfully."
}Use the actual response from your legacy client to decide what to do next.
Check that you are using your existing legacy SSE/API key configuration. Connecting the public ChatGPT plugin does not make convert_audio_to_midi available.
Check that the URL points to the audio file and remains accessible to the service. A link that works only inside your signed-in browser may not be usable by the conversion service.
Inspect the task or error information in your existing client workflow. A submitted request is not proof of a completed conversion. Use the actual returned download link once a result is available.
Review the MIDI alongside the source audio in your editor. Dense mixes and overlapping vocals may need manual corrections; changing quantize affects timing rather than guaranteeing pitch accuracy.
These answers apply only to the existing SSE/API key integration, not the public Melogen Music Intelligence ChatGPT plugin.
It accepts song and mix audio, including polyphonic material. Review the resulting MIDI, especially for dense arrangements.
Limits depend on the legacy service configuration and the client that calls it. Supply a URL that the legacy service can reach.
Audio containing vocals can be submitted, but the MIDI result should be reviewed before use, particularly when vocals and instruments overlap.
When supported by the legacy service, save_stems requests separated instrument stems alongside the conversion workflow.
For existing SSE/API key clients. This integration is not included in the public Melogen Music Intelligence ChatGPT plugin.