Cut a 3-minute Short from a 25-minute recording on Sume

Import the recording, trim three beats with video trim, detach the audio, then render all three on Timeline 1.0 using audio.parts[] for the gapless voice. The whole edit of a 25-minute source into a 180-second Short costs $0.38 on Sume. YouTube's help page lets a Short run up to three minutes (read 2026-10-08), so the target length is the maximum.

Why detach twice

Video trim accepts sources up to 1,800 seconds and produces up to 900. Audio detach has the same source cap and a 900 second output cap, so a 25-minute (1,500 second) track needs two range calls. Each detach is $0.01.

curl -X POST https://api.sume.com/v1/audio-detach \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: rec-detach-a" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/artf_demo/rec.mp4",
    "range": { "start": 0, "end": 900 }
  }'

Trim each beat, then join

A trim needs start plus one of end or duration, and it returns a new MP4 you place in video[] with source_in 0. Three beats of 60 seconds make the 180 second Short.

curl -X POST https://api.sume.com/v1/video-trim \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: rec-beat-1" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/artf_demo/rec.mp4",
    "start": 312.0,
    "duration": 60,
    "audio": "drop"
  }'

Cost of the edit

Things that trip people

• All inputs must be media.sume.com files in your workspace; import the recording first with POST /v1/media-imports.
• audio.parts[] takes up to 20 slices and joins them in the sample domain, so the voice has no seam at the cuts.
• Keep the recording to the files you have the right to reuse.

Choosing the three beats

Use a transcript or your own notes to pick beats that stand alone: a claim, a story and a call to action each need to make sense without the minutes around them. Cut on sentence ends, since the audio parts are joined sample-exact and a cut mid-word will be audible. Because trim re-encodes frame-accurately by default, the start you ask for is the start you get; keyframe precision is faster but can begin up to one GOP early, so re-base against actual_start_seconds if you choose it. Keep the original recording in the workspace in case you want a different cut tomorrow.