Seedance Multi Shot Prompt Guide for Seedance 2.0
How to write a Seedance multi shot prompt for 2.0: shot labels, one camera move per shot, cut markers, stable character labels, and a five-shot template to run.
By OceanToken Team

Editor's note: Most bad multi-shot Seedance clips come from bad shot lists, not a bad model. People type a paragraph of plot and hope the model invents coverage. Write the prompt the way a script supervisor would: numbered shots, one job each. Our default is four to five shots in a 10 to 15 second clip, and no more.
Short answer: A good Seedance multi shot prompt labels each shot in order (Shot 1, Shot 2), puts the camera move first, gives one action per shot, and marks the transition. Keep it to roughly 60 to 100 words. Name the character once with a few fixed features and reuse that label. Expect the model to choose where each cut lands.
Why multi-shot is a different skill
A single-shot prompt describes a moment. A multi-shot prompt describes coverage: a wide shot to set the scene, a medium shot for the action, a close-up to land the detail.
Runway's Seedance 2.0 guide says the model generates up to 15 seconds "across one or several shots." That is enough for a few shots, not a full scene. Plan for a beat, not a story.
The label format that works
Use plain labels: "Shot 1:", "Shot 2:", in the order events happen. Inside each shot, Runway suggests a fixed order:
- Camera move or transition
- Subject action and expression
- Change in space
- Sound
Put the transition at the start of the shot, as in "Cut to a close-up of..." Runway also warns that "one long undivided paragraph gives the model nothing to pace against," so cuts may land where you did not plan them.
One action and one camera move per shot
Runway states it directly: asking for "push, pull, pan, and tilt at once increases image instability." A single move per shot holds steadier.
Heyuan110's guide gives the same rule. If you want two camera moves, "that's two shots." The same applies to action. An "and then" inside a shot usually means you need another shot.
Fast action is a risk area too. Runway notes that sprinting, large jumps and violent motion are where warping tends to appear.
How to stop character drift between shots
Drift is when your courier's jacket changes color in shot 3. Three fixes come from the guides:
- Fixed label. Runway says to define the subject with "two or three stable static features" and use the same label every time.
- Image reference. Heyuan110 notes that identity drifts in text and images anchor it. Give every reference a stated job, such as "@Image1 = the courier's appearance."
- Fewer assets. Runway says four to five references tend to work better than filling the limit.
A Seedance multi shot prompt template to copy
Our take: Use this structure for every multi-shot Seedance prompt. Style line, character block, labeled shots, audio, constraints.
Style: warm dusk, 35mm film grain, muted teal and orange.
Character: the courier, yellow jacket, gray helmet, red bike.
Shot 1: Wide locked-off shot. The courier rides down an empty cobbled street.
Shot 2: Cut to a lateral tracking shot at wheel height. Tires splash a puddle.
Shot 3: Cut to a close-up. The courier's gloved hand rings the bell.
Shot 4: Cut to a slow push-in. A door opens and a child smiles.
Shot 5: Cut to a slow pull-back. The courier waves and rides off.
Audio: (no music) <street ambience, bike chain ticking>
Constraints: no subtitles, no logos, no watermark.
Runway's symbol set uses parentheses for music and angle brackets for sound effects. It also says unwanted subtitles and logos show up unless you add constraints at the end.
Do hard cuts really happen?
The guides disagree. Heyuan110 says Seedance 2.0 "handles multi-shot natively with explicit 'cut to' markers." MindStudio says it "doesn't produce hard cuts the way editing software does," and suggests generating shots separately for true cuts.
Runway adds that precise per-shot timing "is unstable." So treat a hard cut inside one generation as a bonus, not a contract. Write every prompt so the clip still works if two shots blend.
How many shots fit in one clip?
MindStudio suggests two to three beats for a 5-second clip and three to four for 10 seconds. For 15 seconds, we cap at five shots. Past that, shots get too short to read.
If one shot keeps failing, change one thing at a time. Heyuan110 reports that single-variable changes converge in three to four rounds.
How to test this on your workload
- Run the template five times at 720p, 15 seconds.
- Count how many runs show five distinct shots.
- Check whether the jacket, helmet and bike stay the same across shots.
- Remove the "Cut to" markers and run five more. Compare.
Running it on OceanToken
On OceanToken, Seedance 2.0 runs as bytedance/seedance-2.0-upscaled (official render plus our upscaler, 720p or 1080p) or bytedance/seedance-2.0 (Native, sent unchanged to ByteDance's official API). The /seedance-api page recommends Native for scenes with several people, hands or close-up faces. Use base URL https://api.oceantoken.ai/v1. See the current rate on /seedance-api.
FAQ
Does Seedance 2.0 do hard cuts? Sometimes. Guides disagree, so test with and without "Cut to" markers on your prompt.
How many shots fit in 15 seconds? Four to five is a safe ceiling.
Should I give each shot a duration? Usually not. Runway says forced timing is unstable and can give unpredictable results.
How do I keep a character consistent? Reuse one label with two or three fixed features, and add an image reference with a stated job.
Can I skip labels and write prose? You can, but you lose control over where cuts land.
Sources
- Runway: Seedance 2.0 prompt guide (checked 2026-11-02)
- Heyuan110: Seedance 2 prompt guide (checked 2026-11-02)
- MindStudio: timeline prompting for Seedance 2 (checked 2026-11-02)
- OceanToken Seedance API page (checked 2026-11-02)
Get your OceanToken API key
- Create an OceanToken account and an API key.
- Copy the snippet below, paste your key, and run it.
- Check the current rate on /seedance-api.
curl https://api.oceantoken.ai/v1/videos \
-H "Authorization: Bearer $OCEANTOKEN_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "bytedance/seedance-2.0-upscaled", "resolution": "720p", "duration": 15, "prompt": "Shot 1: Wide locked-off shot. A courier rides down a cobbled street. Shot 2: Cut to a close-up. The courier rings the bell."}'
More from the blog
A Unified API for Chat, Image and Video
Building on a unified API for chat, image and video: sync vs async calls, billing units, retries, timeouts and storage, with three calls and a launch checklist.
Kimi K3 API International Access Guide
Use the Kimi K3 API internationally: sign up with email, pay by card, create one key and call moonshotai/kimi-k3 with an OpenAI SDK. Steps, code and checks.
Call Tencent Hy3 With the OpenAI SDK
Call Tencent Hy3 with the OpenAI SDK outside China: base URL, model id, streaming, tool calls, reasoning_effort values, Hy3 vs Hy3 preview, and token billing.