How to Extend AI Video Clips with Wan 2.6

On Astorie's canvas, route the source clip into a frame-extraction tool node. Scrub to the final stable frame — the moment before any motion accent or fade. This frame becomes the seed for Wan 2.6's extension. The cleaner the seed frame, the cleaner the extension joins.

Pipe the extracted frame into a Wan 2.6 image-to-video node. Wan generates 4-8 second clips from a starting frame. The output continues from the seed frame, so when stitched after the source, the cut reads as continuous if the prompt matches the original motion direction.

The prompt should restate the source's shared visual language and add the new motion: "character continues walking forward through autumn forest, soft golden key light, same wardrobe, slow handheld camera." The repetition ("same lighting, same wardrobe") is what holds the extension cleanly to the source.

Wan 2.6 base handles general I2V extension. For tighter character continuity, route through Wan Animate Move (motion reference) — feed the source clip's tail motion as the motion reference and a character anchor as the appearance reference. Animate Move outputs new footage with the same motion vector and the same character. More credits, but tighter control.

On the canvas, send both source and Wan extension into the sequence builder as adjacent clips. The cut should be invisible if the seed frame and the prompt continuity are right. If a small visible jump remains, blend with a 0.5s cross-fade in the NLE export — Wan extensions are typically clean enough that a hard cut works.

Wan's open-weight architecture keeps Sutui costs low. If the first extension does not match cleanly, regenerate at minimal credit cost — try varying the prompt or the seed frame timestamp. For high-volume work (extending 20+ clips for a campaign), Wan beats Pixverse on per-clip cost.