Audio Separation

Split any soundtrack into separate audio layers — Dialogue, Music, Effects, and a without-dialogue (M&E) track. Built for dubbing and localisation, podcast cleanup, re-scoring, and handing an editor usable stems.

This is Kolbo's own pipeline, not a passthrough to a single vendor. Two things make it different from a generic "vocal remover":

• It masks rather than resynthesises. The returned layers sum back to the original at −30 to −37 dB, so the audio you get back is the audio you put in, filtered — not a reconstruction of it.
• A speech classifier decides what "dialogue" means. A vocal mask is not the same thing as speech: engines, impacts and centred ambience land in it at full volume. When a clip has no spoken words, that mask is folded back into Effects instead of being handed to you labelled as Dialogue.

All three endpoints accept an optional project_id body field that bills against and files the work into a specific project. See Projects.

Endpoints

POST /api/v1/audio/separate         # 5 credits  — the main split
POST /api/v1/audio/clean-dialogue   # 17 credits — remove voice leaking into the M&E track
POST /api/v1/audio/ambience         # 17 credits — pull room tone out of the Effects bed

Two ways to name the source

Every endpoint takes exactly one of:

Sending both returns 400. The two paths differ in one important way:

Prefer generation_id whenever the source is a Kolbo generation.

POST /v1/audio/separate

Splits the soundtrack. 5 credits.

Request

*Exactly one of audio_url / generation_id.

curl -X POST https://api.kolbo.ai/api/v1/audio/separate \
  -H "X-API-Key: kolbo_live_..." \
  -H "Content-Type: application/json" \
  -d '{"audio_url": "https://media.kolbo.ai/uploads/interview.mp4"}'

Response

{
  "success": true,
  "source": "url",
  "layers": [
    { "type": "original", "label": "Original",          "url": "https://media.kolbo.ai/.../original-....wav" },
    { "type": "dialogue", "label": "Dialogue",          "url": "https://media.kolbo.ai/.../dialogue-....wav" },
    { "type": "music",    "label": "Music",             "url": "https://media.kolbo.ai/.../music-....wav" },
    { "type": "sfx",      "label": "Effects",           "url": "https://media.kolbo.ai/.../sfx-....wav" },
    { "type": "me",       "label": "Without dialogue",  "url": "https://media.kolbo.ai/.../me-....wav" }
  ],
  "has_dialogue": true,
  "has_music": true,
  "credits_used": 5
}

All layers are 48 kHz stereo WAV on the Kolbo CDN.

POST /v1/audio/clean-dialogue

Strips voices still faintly audible in an M&E track. 17 credits.

Kolbo's separation leaves roughly −33 to −40 dB of speech in the M&E on clean dialogue (inaudible) and about −19 dB on a dense crowd scene (faint). Use this only when that residue actually matters to you — it trades fidelity to do its job, reconstructing the bed at about −12 dB against the −30 dB of the masking split that produced it. It is an escalation, not a step in the normal flow.

*Exactly one.

{
  "success": true,
  "source": "url",
  "layers": [{ "type": "me", "label": "Without dialogue", "url": "https://media.kolbo.ai/.../me-....wav" }],
  "passes": 2,
  "margin_db": -18.4,
  "already_clean": false,
  "credits_used": 17
}

passes is how many removal passes ran before the result stopped improving. already_clean: true means the analysis found nothing worth removing and url is the track you sent.

POST /v1/audio/ambience

Pulls room tone / atmosphere out of an Effects or Music bed, leaving hard effects behind. 17 credits.

*Exactly one of audio_url / generation_id.

{
  "success": true,
  "source": "url",
  "layers": [
    { "type": "sfx",      "label": "Effects",  "url": "https://media.kolbo.ai/.../sfx-....wav" },
    { "type": "ambience", "label": "Ambience", "url": "https://media.kolbo.ai/.../ambience-....wav" }
  ],
  "skipped": false,
  "levels": { "target": -24.1, "residual": -31.7 },
  "credits_used": 17
}

skipped: true means the bed had no separable ambience and layers is empty. As with clean-dialogue, the pass ran and is billed.

Errors

Rate limits and timeouts

• 10 requests/minute per account, shared with the Kolbo web app's own separation surface, so switching between them does not double your allowance.
• The route holds its connection for the whole run and gives up at 10 minutes. Set your HTTP client's timeout above that; a shorter client timeout aborts your side while the server keeps working, and you lose the result.
• Concurrent calls are independent — but each one spends credits, and the audio_url form has no cache. Two identical calls separate twice and bill twice.

Related

• Transcription — if you want the words as text, not the audio layers
• Video Editing — generate_audio adds an AI-generated audio layer; it is not the inverse of this endpoint
• Media Library — POST /v1/media/upload to get a public URL for a local file first