Spotify Clips: 60-90 seconds, one per episode, cut it with Sume

Spotify's creator guidance describes clips as 60 to 90 seconds long, one clip per episode (read 2026-10-02). To make one from a finished video episode with Sume, transcribe it with video inspect, choose one 60 to 90 second window from the sentence segments, and cut that window with video trim. Sume does not upload to Spotify; you publish the file yourself.

What Spotify says about clips

Because Spotify allows one clip per episode, the choice of moment matters more than volume. The article also names the two numbers creators can watch afterwards. It does not list an aspect ratio or file specification and points readers to Spotify's support documentation, so read that before you export.

Find the moment with video inspect

Run video inspect on the episode with transcribe: true and segmentation: { mode: "sentence" }. The probe and stills are unbilled; the transcript is billed per audio minute (the docs list $0.01, confirm live in GET /v1/catalog). The source must be a clip already on your workspace's media.sume.com, so import it first, and it must be at most 1800 seconds, so a longer episode needs a split first.

• Read the sentence segments[] and pick a start at the beginning of a sentence.
• Pick an end at a sentence boundary that lands between 60 and 90 seconds later.
• Inspect returns facts, stills and a transcript, not a ranked best moment. You or your agent decide the window.

Cut the window with video trim

Send start and duration (or end) to POST /v1/video-trim with an Idempotency-Key. Duration may be 0.2 to 900 seconds, so a 60 to 90 second clip is well inside range. The default precision: "exact" is a frame-accurate re-encode. keyframe is a stream copy that can start a GOP early, in which case read actual_start_seconds from the result.

The public rate in the docs is $0.02 per job. Poll GET /v1/jobs/:id/status and read GET /v1/jobs/:id/result for the new video_url. Through hosted MCP the flow is video_trim, then jobs_wait, then jobs_result.

Optional: burn captions, and what is out of scope

If the clip needs burned-in captions, video captions takes a public HTTPS video URL and reserves $0.20 for videos up to 60 seconds under the current fixed estimate. The docs give no price for 61 to 90 seconds, so check the catalog before you batch longer clips.

Sume does not publish to Spotify, and it does not know which moment will perform. Treat the Spotify analytics as your feedback loop and pick the next episode's clip the same way.