How-to + prompt method + same-plan proof

Facts checked August 19, 2026

How to use Seedance 2.0: turn an idea into a testable prompt before generating

To use Seedance 2.0, turn the idea into seven observable fields: subject, visible action and end state, camera, timing, light, sound and constraints. Load them in Shot Planner, inspect the composed prompt, then explicitly choose the supported 18-credit Pro Fast draft or 122-credit Seedance 2.0 Fast path. Nothing generates automatically.

Prompt structure

7 fields

Subject through constraints

Same-plan runs

2

Internal owner canaries, not a success rate

Starter draft

18 credits

Seedance 1.0 Pro Fast · 720p · 5s

2.0 Fast

122 credits

720p · 5s · audio requested

How to use Seedance 2.0 in five controlled steps

Plan the shot before opening a paid model. The same structured fields can first prefill the supported 18-credit Pro Fast draft, then move to the 122-credit Seedance 2.0 Fast path only when its interaction or audio capability is the reason for upgrading. Neither handoff generates automatically.

1. Turn the idea into observable fields

Name the subject, visible action and end state, one camera move, timing, light, sound and constraints.

2. Load the worked example in Shot Planner

Open the allowlisted template, review every field and change only what the shot actually needs.

3. Inspect the composed prompt and checks

Confirm that the final prompt and expected visual checks describe the same short take before choosing a model.

4. Prefill the supported starter draft

Open Seedance 1.0 Pro Fast at 18 credits to test the plan; the handoff never submits automatically.

5. Choose Seedance 2.0 explicitly when needed

Use 2.0 Fast at 122 credits only when the plan needs its stronger interaction or audio path, then review the shown charge again before generating.

The seven-field Seedance prompt formula

Write the shot in this order: subject and setting → visible action and end state → one camera move → timed beats → motivated light → sound or dialogue → explicit constraints. The order makes every instruction inspectable, but no prompt structure guarantees that a model will obey every field.

  • Subject and setting: define only what must stay recognizable in frame.
  • Action and end state: say what visibly changes by the final beat.
  • Camera: use one dominant move for a five-second diagnostic take.
  • Timing: assign the opening, middle and final beat when order matters.
  • Light: name source and direction instead of using “cinematic” alone.
  • Sound: quote exact dialogue or name ambience; some tiers do not output audio.
  • Constraints: list disallowed text, logos, cuts, extra subjects or hands.

Worked example: idea → Shot Plan → final prompt → checks

Idea: a small ceramic automaton folds one blue sheet into a paper crane during one continuous five-second take. This is a company-authored diagnostic shot, not a promise that a short take can reliably complete every fold or camera instruction.

Both provider submissions used the same composed prompt (SHA-256 31d491f7…de417), five-second duration and 16:9 aspect ratio. Only the requested model path differed.

FieldFrozen instruction
SubjectA small unbranded ceramic automaton at a clean walnut workbench with one centered sheet of blue paper.
ActionFold the single sheet step by step and leave a complete crane upright by the final beat.
CameraMedium close-up, one continuous slow clockwise orbit, no cut.
Timing0–1s meet paper; 1–4s deliberate folds; 4–5s release the crane.
Light / soundWarm window light; paper folds, mechanical tick and room ambience; no dialogue.
ConstraintsNo text, captions, logos, watermark, extra hands, extra paper, cut or recognizable character.

Copyable final prompt and expected visual checks

A small unbranded ceramic automaton at a clean walnut workbench, one sheet of blue paper centered in front of it. The automaton folds the single sheet step by step; by the final beat the paper is a complete crane resting upright on the bench. Camera: Medium close-up, one continuous slow clockwise orbit; no cut. Timing: 0–1s hands meet paper; 1–4s deliberate folds; 4–5s hands release the completed crane. Light: Warm morning window light, restrained contrast, physically plausible ceramic and paper reflections. Sound: Soft paper folds, quiet mechanical tick, natural room ambience; no dialogue. Constraints: No text, captions, logos, watermark, extra hands, extra paper, scene cut, brand marks or recognizable character design. Duration: about 5 seconds. Aspect ratio: 16:9

  • One blue sheet remains present throughout the take; no extra paper appears.
  • The fold has visible intermediate motion and ends as a recognizable upright crane.
  • The ceramic automaton remains the only acting subject; no human hands or arms appear.
  • The shot stays continuous with no cut, text, caption, logo or watermark.
  • A single five-second result can reveal instruction failures, but cannot establish average quality, reliability or a success rate.

Same plan, two current provider paths

These are first-party internal owner canaries from August 19, 2026. They are two observations, not an average, benchmark or claim that one tier always wins.

Starter path: Pro Fast redirect probe

The request used the old Lite slug, which fal now marks deprecated and redirects to Seedance 1.0 Pro Fast. The 5.04-second H.264 output ended with a blue paper crane, but introduced large human hands and arms despite the “no extra hands” constraint. It had no audio stream. This is failure evidence, not a Lite result.

Request 01a019ba-2c13-77c3-8b4e-8dde0d555f8b · 30.8s observed · 18 site credits configured · provider invoice cost not yet joined

Upgrade path: Seedance 2.0 Fast

The 5.09-second 1280×720 H.264/AAC output kept the automaton as the only acting subject, visibly folded the blue paper and ended on a crane. Sampled frames did not prove the requested continuous orbit, so that field remains unconfirmed rather than counted as a success.

Request 01a019ba-a436-70e3-8c4d-5622cffaf43e · 142.1s observed · 122 site credits configured · $1.21968 configured cost anchor, not invoice truth

Why this guide no longer calls the starter path “Lite”

fal currently marks the former Seedance 1.0 Lite endpoint as deprecated and says requests redirect to Seedance 1.0 Pro Fast with changed pricing. First Take Lab therefore calls the supported Pro Fast endpoint directly. The site still charges 18 credits for this draft; 25 starter credits cover one run.

  • Current official Pro Fast token pricing implies about $0.108 for a requested 720p, 24fps, five-second run; invoice-level cost is still unverified.
  • The old $0.18 Lite cost anchor is no longer used as current provider truth.
  • The two outputs differ in encode dimensions and audio stream; they are not quality-equivalent tiers.
  • A completed provider task does not mean every prompt field succeeded.

Use the observed state to choose the next action

Observed stateWhat it meansNext action
Blocked before submitThe request did not pass a local safety or product gate.Remove the risky instruction; no provider call should occur.
Insufficient creditsThe balance is below the model charge shown before submit.Return to the 18-credit Pro Fast draft or cover only the exact gap.
Provider failedThe task reached the provider but produced no valid video.Confirm the recorded failure and credit return before simplifying the plan.
Completed but an instruction missedCompletion proves a file exists, not that the shot followed every field.Compare against the visual checks and change one field before another run.

Turn the method into an executable plan

Load the exact seven fields into Shot Planner, review or change them, then explicitly prefill the supported 18-credit Pro Fast draft. Loading the plan is free and never generates automatically.

Shot Planner

The allowlisted F2 link preserves source, experiment and template IDs without putting the creative prompt in the landing URL.

Load the same Shot Plan
Free and cost truth

See exactly what 25 starter credits can fund and why Seedance 2.0 remains a paid upgrade.

Check the cost boundary

Frequently asked questions

What is a good Seedance prompt formula?

Use subject and setting, visible action and end state, one camera move, timed beats, motivated light, sound or dialogue, and explicit constraints.

How do I use Seedance 2.0 on First Take Lab?

Structure the idea in Shot Planner, inspect the final prompt, test the plan with the supported 18-credit Pro Fast draft if appropriate, then explicitly select Seedance 2.0 Fast at 122 credits when its stronger interaction or audio path is needed.

Where can I access Seedance 2.0?

First Take Lab exposes the verified five-second 720p text-to-video path in the browser. You can prepare a plan without signing in; authentication, safety checks, rate limits and the displayed credit check run before any provider submission.

Does a structured prompt guarantee a usable result?

No. In this two-run sample, the lower-cost path completed but violated the no-extra-hands constraint. A structured plan makes failures easier to inspect; it does not remove model uncertainty.

Is Seedance 1.0 Lite still available here?

No current output is labeled Lite. fal marks that endpoint deprecated and redirects it to Seedance 1.0 Pro Fast, so First Take Lab now requests Pro Fast directly.

Will opening Shot Planner spend credits?

No. The planner composes and stores the plan locally. It only prefills the generator, and generation still requires an explicit click.

Both videos are First Take Lab internal owner canaries using company-authored, unbranded input. No customer media, face, logo or third-party brand was supplied. The samples prove only these two runs; they do not establish average quality, success rate, external first value or invoice-level cost.