The one move: video-to-video
The one move: video-to-video
Higgsfield Academy
Caption
Everything in this course is the same move, repeated. You hand Seedance 2.0 a clip you actually shot, tell it in one sentence what to change, and it hands the clip back — your motion, your face, your camera move untouched, and only the part you asked for repainted. That is video-to-video, and it is the entire workflow.
One clip in, the same shot out
You are not rebuilding your footage. The real clip goes in as a video input; Seedance keeps everything that makes the shot yours — your walk, your timing, the way your jacket moves — and morphs only what you named. No keyframing, no masking, no tracking, no rotoscoping. Real shot in, transformed shot out.
Always keep an anchor
It reads as real because something in frame stays genuinely real. Your hands on the wheel, your face, the dashboard close to camera — that is your anchor. Change everything at once and the brain smells fake; hold one real anchor and it buys the rest. Every lesson here protects an anchor and changes one thing around it.
Why native 4K, not upscaled
The results come back in native 4K — four times the resolution, not an upscale bolted on afterward. That is what lets you punch into a generation in the edit and still hold detail. You generate once and pull two or three sharp shots out of it.
Three levels, hardest last
Level 1 — Swap the world
Change the whole environment around you: desert, neon city, lava, above the clouds.
Level 2 — Change an element
Transform one thing in frame: head on fire, a hand into a snake, creatures on a wall.
Level 3 — Handheld showcase
Full moving-camera cinema: a wing-walker, a jungle temple, sauropods, a kraken storm.
A free Seedance prompting skill writes these keep-lists for you — the prompt in every lesson is what it produced. Copy any of them straight from the player and change one thing at a time.