How should I structure a Seedance 2.0 prompt?
State subject and setting, visible action, one main camera move, timing, light, sound and constraints in that order.
Prompt library
Facts checked August 12, 2026
Seedance 2.0 prompts that specify the shot — not just the picture
Use the formula Subject + Action + Camera + Timing + Light + Sound + Constraints. Five recipes now have paid provider outputs; the macro product test remains visibly pending in the provider queue.
Prompt parts
7
Subject through constraints
Completed output
6
Five recipes; first tested on two tiers
Original recipes
6
Test status shown per recipe
Current path
720p · 5s
Text to video with audio
The Seedance 2.0 prompt formula
Subject and setting → visible action → camera move → timing or sequence → light and style → sound or dialogue → explicit constraints.
- Subject: include only details the model must preserve.
- Action: use verbs and an end state, not mood alone.
- Camera: choose one dominant move for a five-second test.
- Timing: say what happens first, next and last when order matters.
- Sound: put exact speech in quotation marks; otherwise name ambience and effects.
- Constraints: state no captions, logos, cuts or extra subjects when needed.
Six original Seedance 2.0 prompt recipes
Ceramic automaton folds an origami crane
A small unbranded ceramic automaton sits at a clean wooden workbench and carefully folds one sheet of blue paper into a crane. Medium close-up, one continuous slow clockwise camera orbit, warm morning window light, physically accurate finger and paper movement. Natural room ambience, soft paper folds and a quiet mechanical tick. No dialogue, no captions, no logos, no watermark.
Paid canary tested on Fast and Standard
Use this prompt in FastTwo dancers exchange momentum
Two contemporary dancers cross a bare rehearsal studio and exchange momentum through one hand-to-hand turn, then stop in balanced opposing poses. Full-body wide shot, slow lateral tracking camera, overcast skylight, realistic weight transfer and foot contact. Shoe movement and room ambience only. No text, no logos, no extra people.
Paid Fast canary completed in 151.4 seconds · 121 credits
Use this prompt in FastBarista dialogue with a single rack focus
A barista places a cup on a quiet café counter and says, “Oat latte for Maya.” The waiting customer smiles and replies, “Perfect, thank you.” Medium two-shot, one rack focus from cup to customer, warm practical lights, natural lip movement, cup and café ambience. No subtitles, no brand marks, no camera cuts.
Paid Fast canary completed in 151.3 seconds · 121 credits
Use this prompt in FastMacro product material test
An unbranded brushed-steel travel flask rotates slowly on a dark stone pedestal as a narrow ribbon of soft light moves across its surface. 100mm macro lens, slow push-in, realistic reflections and condensation, black studio background. Quiet room tone. No labels, no hands, no text, no watermark.
Paid Fast canary still in provider queue; no result claimed
Use this prompt in FastPaper boat crosses changing weather
A blue paper boat floats along a shallow curb stream as light rain begins, passes under fallen leaves, and enters a patch of late-afternoon sunlight. Low tracking shot at water level, continuous movement, realistic paper absorption and small ripples. Rain and water audio only. No people, no captions, no cuts.
Paid Fast canary completed in 121.0 seconds · 121 credits
Use this prompt in FastMiniature train reveal
A handmade miniature red train exits a paper tunnel, crosses a wooden bridge, and reveals a tiny mountain village at sunrise. Begin with a close tracking shot beside the train, then gently crane upward for the reveal, physically consistent wheels and steam. Soft mechanical sound and morning birds. No logos, no text, no abrupt cuts.
Paid Fast canary completed in 151.0 seconds · 121 credits
Use this prompt in FastWhat the paid-canary prompt actually produced
Both runs returned a 1280×720, 5.04-second MP4 with audio enabled. Fast completed in 124.9 seconds at an expected $1.2095 provider cost; Standard completed in 222.0 seconds at an expected $1.5170 provider cost. The two outputs interpreted the automaton differently, which is useful evidence: the prompt controlled the action and setting, not exact character design.
| Tier | Observed result | Cost anchor |
|---|---|---|
| Fast | Robot torso and hands framed tightly while folding white paper. | 121 credits |
| Standard | Smaller mannequin-like automaton with clearly blue paper. | 152 credits |
Prompt mistakes that waste expensive takes
- Describing a beautiful still image without saying what changes during the video.
- Stacking several camera moves into five seconds without an order.
- Repeating source-image details instead of describing motion.
- Using vague quality claims such as “cinematic masterpiece” instead of observable light and lens cues.
- Changing subject, action, camera and style at once after a miss, so the next result teaches nothing.
Frequently asked questions
Do long prompts always work better?
No. Add detail only when it controls something visible or audible. Contradictory detail makes the model solve too many instructions at once.
How do I prompt dialogue?
Put the exact spoken line in quotation marks, identify the speaker and keep the shot simple enough for five seconds.
Are all six recipes tested?
Five recipes have completed paid outputs, including Fast and Standard outputs for the automaton recipe. The macro product recipe was submitted but remains in the provider queue, so this page does not claim a result for it.
No stock gallery output is presented as a First Take Lab result. Every embedded video is tied to a dated paid canary, and the provider-queued recipe remains explicitly labeled without a claimed result.