Clips for a 180-second Short: Wan 3.0, Seedance 2.5, Kling 3, Omni

YouTube says Shorts can be up to 3 minutes long. At 180 seconds, a model that makes 30-second clips needs 6 of them, one that stops at 15 seconds needs 12, and Gemini Omni Flash at 10 seconds needs 18. Join them with Timeline 1.0, whose render.strategy auto chunks past 12 segments.

The YouTube side

"Get started creating YouTube Shorts" (read 2026-10-05) says Shorts can be up to 3 minutes long and that the maximum upload resolution is 1080p. It also says to upload vertical videos. The math below is about your side: how many generated clips make 180 seconds.

Clip count by model

The numbers use each model's maximum duration from the Sume docs, so they are the minimum count. In practice you pick shorter clips for a better take and need more.

Joining them

Timeline 1.0 accepts 1 to 200 video[] slots and an audio spine up to 1800 seconds. With 18 slots, the docs say render.strategy auto chunks past 12 segments, so you do not set it. Keep transitions short; each is at most 1 second and half the shorter neighbour.

Model clips may run at different frame rates. Set output.fps to 24, 25, 30 or 60, or you get the output_fps_resamples_sources warning.

curl -s -X POST https://api.sume.com/v1/timeline-1.0/plan \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d @short-180.json

Cost shape

Timeline is $0.10 per ceil output minute in the docs, so a 180 second render is three billed minutes at that list rate. Each clip is its own model job on top; read per-second rates from GET /v1/catalog, since they change.

Fewer, longer clips mean fewer joins but less choice in each take. Six 30-second clips are cheaper to manage; eighteen 10-second clips give more cut points.

• Use the plan route before the render.
• Keep clip order in a manifest.
• Add the music bed as soundtrack with duck_db if a voice is present.