FLUX 3 video continuation caps at 15 s: extending a clip on Sume

Black Forest Labs' release notes say the maximum duration for FLUX 3 video continuation (v2v) requests is now 15 seconds. Sume's video docs do not describe a continuation mode. To extend a clip on Sume you start a new clip from a frame or from reference videos, and each request is validated against the chosen model's own duration range.

BFL facts are from its release notes dated August 17, 2026; Sume facts are from the video input modes and video generation docs, read 2026-10-01.

What changed in FLUX 3 video continuation?

BFL lists the limit as 15 seconds for v2v. Text-to-video, image-to-video, and draft enhance are unchanged at up to 20 seconds.

What input modes does Sume list instead?

The Sume video docs list four ways to start a request, and none is named continuation.

How do I extend a clip with a last frame?

Pull a frame from the end of the first clip with video frames, then pass it as image_url on the next request. Frame extraction takes a media.sume.com clip and at[] times, and the docs say it is unbilled. The docs only say that video requests need public HTTPS media URLs, so check that the frame URL you send is reachable by the model request. The step-by-step version is in extend an AI video past 30 seconds.

Watch the join: nothing in the docs promises that two clips match in motion, so review the seam before you publish.

curl -X POST https://api.sume.com/v1/video-frames \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: clip-1-last-frame" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/artf_demo/clip-1.mp4",
    "at": [9.5]
  }'

What duration can I request on Sume?

duration is an integer number of seconds validated against the model you picked, so read the models endpoint before you pin a value. The docs give examples: seedance-2.5 accepts 4 to 30 seconds and wan-3.0 accepts 2 to 30 seconds.