ByteDance's Seedance 2.5 prompt guide (18 Aug 2026) publishes a full template. This sheet is that template, with the rules around it quoted.
- The plain answer: write the shot in ByteDance’s order — length and framing first, the action third, the things to avoid last — and keep each beat of the action to about 6–8 seconds.
- Up to 30 seconds in one go. Longer videos come from extending a clip; ByteDance says you can extend twice, and its Dreamina app advertises a 180-second beta mode.
- Eight parts: Format → Subject → Action → Environment → Camera → Look → Timing → Audio & Constraints.
- Format first, action third. Duration, aspect ratio and shot structure come before anything else.
- Bind references by name — ByteDance's example: “The same woman from @Image1 wearing the red jacket”.
- One camera behaviour per shot. ByteDance: “Give every shot one camera behavior and every sequence one dominant visual rule.”
- Beats of 6–8 seconds. ByteDance calls that “usually a solid default”.
- Exclusions last, short and concrete, after a positive Preserve clause.
Every rule on this sheet is quoted from the vendor's own published documentation, linked where it is used.
The formula
Dreamina, Seedance 2.5 prompt guide, 18 Aug 2026, read at source 16 Sep 2026.
One generation makes up to 30 seconds. ByteDance: “Create videos up to 30 seconds in a single generation, with the option to extend twice”. Extending means asking the model to add more video onto the end of a clip you already have. Dreamina, ByteDance’s consumer app, advertises more: “30 seconds in standard mode or extend them to 180 seconds with the beta long-video mode”. Both are vendor claims.
ByteDance Seed, Seedance 2.5, no date on the page, read at source 1 Oct 2026; Dreamina (ByteDance), Seedance 2.5, no date on the page, read at source 1 Oct 2026.
Template:
[Duration, aspect ratio, shot structure]. Use [@reference] as the consistent [character/product/location]. In [environment and time], the subject [precise action]. Camera: [shot size + movement + lens/focus]. Visual direction: [lighting + palette + texture + realism]. Timing: [beat 1] → [beat 2] → [beat 3] → [resolution]. Audio: [dialogue/ambience/SFX/music]. Preserve [identity, wardrobe, logo, object geometry]. Avoid [unwanted cuts, camera behavior, artifacts, extra objects, text].
Worked example:
18 seconds, 16:9, single continuous take. Use @Image1 as the consistent ceramicist. In a north-facing studio at dusk, she centres a ball of clay and draws it upward. Camera: medium shot, slow push-in, 50mm, focus held on her hands. Visual direction: window light only, muted earth palette, visible clay texture, photoreal. Timing: 00–07s hands settle and the wheel speeds up → 07–14s the form rises → 14–18s she sits back and exhales. Audio: wheel hum, ambient room tone, no music. Preserve identity, apron, wheel geometry. Avoid cuts, camera shake, on-screen text.
Breakdown — what each part does
Format
Duration, aspect ratio, shot structure. First — it frames everything after it.
Subject
Identity, wardrobe, product or hero object. Bind it to a reference here if you have one.
Action
Visible movement with a clear start and finish. ByteDance: “Write observable instructions.”
Environment
Location, time, weather, practical detail.
Camera
Shot size, lens feel, movement, focus — one behaviour per shot.
Look
Lighting, palette, texture, realism level. If output keeps arriving glossy, the fix is usually subtraction rather than more style words — see why AI output looks fake, which is this site's craft reasoning rather than vendor guidance.
Timing
Beat by beat, with arrows. ByteDance: “6–8 seconds per beat is usually a solid default, with 3–4 seconds saved for the resolution.”
Audio & Constraints
Dialogue, ambience, SFX — then Preserve [...], then a short Avoid [...].
Does it take a negative prompt?
negative_prompt field — exclusions live inside the prompt.NODreamina, Seedance 2.5 prompt guide, read at source 16 Sep 2026: “Negative instructions work best when they protect a specific production requirement. Keep them short, concrete, and grouped at the end.”
Say what must survive (Preserve) before what must not happen (Avoid), and keep both in one place at the end.
Protect what matters, then exclude only what your shot needs — short, concrete, last.
What not to do
Many guides, including earlier versions of this site, teach a [MOTION]-first bracket structure. It was tested and it works — but it is not what ByteDance publishes. Use brackets as a checklist so you do not forget motion, identity, camera or duration, then order the finished prompt the way the template above does.
Free or paid? What the free plan allows, who owns the output, and when you need to pay: free vs paid AI image, video and music tools.
Where this model sits among the others: which models take a negative prompt, and what each vendor publishes as its formula.
For the expensive-looking version, see the cinematic sheet.
The checklist
Before you send it:
@Image1-style tags?□Preserve [...] before Avoid [...]?□Dreamina (ByteDance), Seedance 2.5 prompt guide, read at source on 16 Sep 2026 · ByteDance Seed, Seedance 2.5 and Dreamina (ByteDance), Seedance 2.5, both read at source on 1 Oct 2026 for the clip-length box.
Model versions and vendor documentation both move. Re-read the source before relying on a specific number.
Every other sheet in this set: all model cheat sheets.
Related on this site: No Negative Prompt Field? · Sound in AI Video
Free printable AI cheat sheets, and AI news every week.
Sixteen one-page AI guides and checklists in one free PDF, no sign-up. And if you want to keep up, the newsletter covers what changed in AI each week, in plain English, with the sources.
SUBSCRIBE FREE ↗ GET THE CHEAT SHEET PACK