Seedance 2.0 Prompt Library — Reusable Templates & Examples

A reusable template keeps the subject, action, camera, lighting, and style fields in a predictable order. That matters because prompt drift often comes from changing both the shot idea and the prompt structure at the same time. A library lets you isolate what actually improved the result: a better action verb, a better camera cue, or a better reference image.

Product ads, talking-character scenes, fashion shots, cinematic b-roll, and image-to-video continuations each need different emphasis. A product ad prompt cares about material detail, branding, and camera polish. A talking character prompt cares more about identity stability, framing, mouth visibility, and audio timing. Organizing prompts by job type keeps your workflow searchable instead of forcing one master template to do everything.

A useful library stores more than text. Save the references, output notes, duration, camera style, failed variants, and the situation where the prompt worked. The goal is to create a repeatable playbook, not just a gallery of isolated wins. This is especially important when teams hand off projects or revisit a workflow weeks later.

The fastest way to learn what a prompt field actually controls is to change one variable at a time and compare outputs. Pick a working base prompt, then run a series where only the camera instruction changes — dolly in, static wide, handheld close-up — while subject, lighting, and style stay identical. Do the same for style terms, action verbs, or reference images in separate rounds. If you change the camera and the style keyword in the same run, you cannot tell which edit caused the difference. Keep a simple log: base prompt, variable changed, new value, and a one-line note on what shifted in the output. Over a few rounds this builds a reliable map of which fields have the most leverage for your specific use case. Teams that skip controlled testing usually waste more generation credits guessing than teams that spend a few runs isolating variables first. The log also helps when onboarding a new collaborator — they can read what was already tested instead of repeating experiments.

Image-to-video prompts work differently from pure text-to-video because the reference image already carries composition, color, and subject identity. Your prompt should focus on what the image cannot convey: motion direction, camera behavior, timing, and sound. If the reference is a product shot on white, the prompt adds the camera arc and lighting shift — not the product description the image already shows. Use the @Image1 tag to assign a specific role: first frame, character reference, style board, or environment plate. A product-shot template might read '@Image1 as first frame, slow 180-degree orbit, soft studio rim light, premium reflections, 5 seconds.' A character-scene template might read '@Image1 as character reference, medium shot, walking forward, gentle handheld sway, warm daylight.' Common mistakes include over-describing what is already visible in the image — this can conflict with the visual signal and cause drift. Start minimal, then add motion and camera cues only when the default output misses your intent.