SeeDance is ByteDance’s video-generation line, built for turning stills into film-grade motion. Four moves in its repertoire:
Upload a still as the first frame; SeeDance animates the scene with natural motion guided by your text — 5–12 second cinematic clips.
Give a start image and an end image; the model interpolates a smooth A-to-B transition — the controlled-transformation move.
Camera movement, subject motion, lighting and mood all steer from the prompt.
5, 10 or 12-second outputs depending on model — longer takes capture more scene evolution.
Animate or restyle existing footage with rich reference control: up to 9 reference images, 3 reference videos and 3 reference audio tracks, plus auto audio generation.
Advanced image-to-video with endpoint control — first frame required, last frame optional — and prompts up to 2,000 characters.
Streamlined image-to-video: first frame plus a text prompt, no advanced controls — the straightforward single-image move.
Pure text-to-video — describe the whole scene in the prompt and generate from scratch, no image needed.
SeeDance 1.5 Pro is the recommended lead; 1 Pro variants for simpler moves; Seedance 2 for reference-driven restyling.
For 1.5 Pro: first frame required (JPG/PNG, <25MB), last frame optional. Text-to-video skips this step.
Describe motion, camera, mood, lighting and environmental change — up to 2,000 characters on 1.5 Pro.
Pick duration (5/10/12s) and ratio where supported. Generation takes ~30 seconds to a few minutes; download as MP4.
“The camera pushes in as she turns” beats “a nice scene of a woman” — SeeDance follows action language.
Name the move: push, pull, orbit, track. One camera intention per clip stays smooth.
“Golden hour, soft haze” carries the grade; contrast changes mid-clip read as flicker.
When the final pose matters, use 1.5 Pro’s last-frame input rather than describing the ending.
Clips run 5–12 seconds (15s on Seedance 2) — stitch takes for longer sequences.
Thirty seconds to several minutes per generation depending on load and settings.
Many interacting subjects can blur or merge mid-motion — feature one or two dancers per frame.
Identical inputs produce sibling clips; save the take you like.
5 · 6 · 7 · 8
One still, one prompt, and SeeDance choreographs the motion — film-grade, twelve seconds at a time.
Generate with SeeDance — free