Ingest Audio

​

Ingest Audio

• /trace/v1/ingest/type1 (Forces the legacy model)
• /trace/v1/ingest/type2 (Forces the new efficient model)

​

Request Body Fields

​

Valid tasks Options:

• full_audio_latent: Encodes full audio sequence latents.
• stem_vocals: Splits vocals stem.
• stem_bass: Splits bass stem.
• stem_drums: Splits drums stem.
• stem_other: Splits accompaniment/other stem.
• light_stem: Splits 2 stems (vocals + accompaniment).
• motif: Computes melody contours / motif chroma.
• section: Runs structural section boundary segmentation.

​

Ingestion In-App Pricing

​

Request Example

{
  "dataset_name": "my-reference-catalog",
  "audio_blobs": [
    "dataset/123e4567-e89b-12d3-a456-426614174000_song_candidate_1.wav"
  ],
  "tasks": [
    "full_audio_latent",
    "light_stem",
    "motif"
  ],
  "webhook_url": "https://my-service.com/webhooks/trace"
}

​

Response

{
  "success": true,
  "data": {
    "job_id": "ingest_job_7e8f9a0b-c1d2-3e4f-5a6b-7c8d9e0f1a2b",
    "cost_usd": 0.0513
  }
}