audio_parts_channel_mismatch: concat needs one channel layout

audio_parts_channel_mismatch means the parts of a timeline audio concat do not share one channel layout, for example one mono file and one stereo file. The docs mark it as a worker refusal, so it can surface after the job is accepted. Make every part the same layout, then re-run.

Where does the error come from?

From the Timeline audio docs, the neighbouring refusals are:

How do I make the layouts match?

When parts come from videos, audio detach has a channels field: source (default) or mono. Send channels: "mono" on every detach that feeds the concat and those parts share one layout (mono). Parts from other Sume audio should be checked for layout before joining; the docs do not list the layout of each generator's output.

curl -X POST https://api.sume.com/v1/audio-detach \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: detach-mono-a" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/artf_demo/a.mp4",
    "channels": "mono"
  }'

Will a retry cost again?

Each timeline audio job is $0.01 flat, so a corrected concat is a new job. The docs do not state how a refused job is billed, so do not assume either way.

How does the job run?

Timeline jobs default to mode: "async". Pass mode: "sync" to wait up to 30 seconds for a 200 finished job, or you get 202 and poll GET /v1/jobs/:id/status and GET /v1/jobs/:id/result; there is no separate GET for the audio or render job. Idempotency-Key is required on the render and audio jobs, and every URL must already be this workspace's media.sume.com audio or video. The channel check runs on the worker, after admit.

Hosted MCP has timeline_audio and timeline_create; the flow is the tool, then jobs_wait, then the result tool. Details are in the Timeline audio docs and Timeline 1.0 docs.