Veo 3.1 extend video, and how to chain clips on Sume

Veo 3.1 extends a video you already made with Veo by 7 seconds, up to 20 times, and only Veo-generated videos qualify. Sume does not list Veo or an extend call, so the way to make a long shot there is to pick a model with a longer limit (seedance-2.5 and wan-3.0 go to 30 seconds) or to chain clips: take the last frame of clip one and send it as the first_frame of clip two.

Google's numbers are from its Veo page; Sume's from Video generation and Video frames, read 2026-09-29.

What are the limits of Veo 3.1 extension?

Google lists: the input must come from a previous Veo generation, 720p, 9:16 or 16:9, 141 seconds or less; the output combines your input and the new part for up to 148 seconds. Videos are stored for 2 days, and referencing a video for extension resets that timer. Extension is limited to Veo 3.1 and Veo 3.1 Fast, not Lite.

Can Sume extend a video?

No. The video docs list no extend field, and POST /v1/videos makes a new clip each time. For one longer clip, choose by length in the catalog: seedance-2.5 accepts 4 to 30 seconds and wan-3.0 2 to 30 seconds; most other models stop at 15.

How do I chain clips from the last frame?

Extract a still from the end of clip one with POST /v1/video-frames (unbilled; it needs a media.sume.com clip and a time in at[]), then send that still as frame_images first_frame for the next clip with a prompt that continues the action.

curl -X POST https://api.sume.com/v1/video-frames \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: frames-clip1-end" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/artf_demo/clip1.mp4",
    "at": [4.9]
  }'

Which approach fits which job?

Chaining can drift: each clip redraws faces and props from a single still, so keep the scene simple.