Smooth Long AI Videos with Auto Chaining

A video over 15 seconds is several shots stitched together. By default the shots render at the same time, each from its own references, so a pose, a light or a face can change a little at a join. Auto chaining is a new option for long videos: each shot opens on the last frame of the shot before it, so the joins are smoother and more consistent. It costs nothing extra. The trade-off is time, because the shots then render one after the other.

What auto chaining does#

• Each shot in the same place opens on the last frame of the shot before it.
• A change of place is a cut: the shot after it does not chain.
• A shot chains only to a shot that passed the quality check, so a bad ending is not carried into the next shot.
• A shot that pins its own first_frame, or sets chain_from_previous: false, does not chain.
• It works on seahorse-* models, in auto, advanced and edit mode.

Sex shots get a reference, not a first frame#

A shot with explicit: true renders its act from anatomy references. A pinned first frame would drop those references, so a sex shot does not open on the previous frame. It gets that frame as one extra reference image instead, and it keeps its anatomy references. See the explicit flag for how to write a shot that renders the act.

Turn it on#

In the Playground, pick a video over 15 seconds (or Advanced mode) and tick Auto chaining. Over the API, add auto_chain: true to the create request:

curl -s -X POST "$NURETA_URL/api/v3/contents/generations/tasks" \
  -H "Authorization: Bearer sk-..." -H "Content-Type: application/json" \
  -d '{"model": "seahorse-480p", "duration": 60, "auto_chain": true,
       "content": [{"type": "text", "text": "a slow evening in her apartment"}]}'

The default is false. A single clip (15 seconds or less) or a direct Seedance model returns 400 InvalidParameter with auto_chain: true, because there is nothing to chain. In advanced mode, every shot after the first chains, except a shot that sets chain_from_previous itself or pins its own first_frame.

In edit mode#

Send auto_chain when you create the draft, or change it before the render with the structured edit (PUT …/plan/structured with auto_chain: true or false at the top level). Omit the field to keep the current setting. The edit re-expands the plan as usual, and the draft then shows scene.autoChain. In the Playground's storyboard editor, the same checkbox sits above the render buttons. See Structured edit.

The render takes longer, the price does not change#

The same seconds render at the same per-second price: auto chaining has no extra charge. What changes is the time. Each shot waits for the one before it, so the total time is close to the sum of the shot renders, not the time of the slowest shot. A 180-second video can take several times longer than without it. You can leave the page; the task keeps running, and a webhook or a poll tells you when it is done. A chained task also gets more time before the automatic timeout refund: 30 minutes more for each shot after the first.

When a chained frame is refused#

If the video model refuses a chained frame, that shot renders again without it, on its own references. The video does not fail because of chaining. A video blocked by the content policy is blocked as before.

Fixing one part of a chained video#

A chained video can be re-rendered part by part, like any other long video (see fix one clip, not the whole video). A re-rendered part chains again, to the part before it as that part is now. If you redo the part before it in the same request, it renders unchained, so two neighbouring parts redone together join with a cut. A part moved to another place does not chain again. The join to the next part can jump, because the next part opened on the old ending. To smooth that join, fix the next part after this fix is done, in a new request. The Playground's confirm step says so when you pick a part like that. An older chained video, made before chained parts could be re-rendered, still cannot be fixed part by part.

For the full reference, see Auto chaining in the docs. For planning a long video, see multi-shot adult scenes up to 3 minutes.