Seedance 2.0 Prompt Generator

What is Seedance?

Seedance is a family of AI video generation models from ByteDance’s Seed research team. It supports text-to-video and image-to-video, and one of the features that made it stand out is the ability to follow prompts that describe several shots in sequence, keeping the subject and style consistent across cuts. Seedance 2.0 is the newer generation and focuses on better motion, prompt following and control. ByteDance makes Seedance available through its own creative apps and through cloud APIs; which version you can use, clip lengths and resolutions depend on the platform, so check the provider you are using for current details.

Because Seedance is good at sequences, the way you organize a prompt matters. A wall of text that mixes two scenes together often produces a muddled shot. Clearly separated shots produce clearer edits.

Prompt structure for Seedance

For a single shot, use the familiar order: subject, action, setting, camera, lighting, style. For a multi-shot sequence, use numbered shots:

1. Start with a one-line overview of style and mood that applies to every shot.
2. Describe the main character once in detail, and reuse that exact description in each shot.
3. Write “Shot 1:”, “Shot 2:” and so on, each with framing (wide, close-up), camera movement and the action.
4. Use explicit transitions such as “cut to” or “match cut to”.
5. Keep the number of shots small. Two or three shots fit a short clip; more tends to rush each one.

Seedance prompt examples

Whimsical fantasy, warm and magical mood, soft cinematic lighting.
Shot 1: Wide shot of a girl of about ten with curly red hair and a green coat pushing open a creaking wooden door in a dark hallway; static camera.
Cut to
Shot 2: Reverse angle, the same girl with curly red hair and a green coat steps into a vast overgrown library, vines on the shelves, floating dust glowing in sunbeams; slow crane up revealing the height of the room.

A ballet dancer in a white leotard performs a slow pirouette on a city rooftop at dawn, her skirt flaring as she spins. Medium-wide shot, slow orbit around her at waist height, the skyline softly out of focus behind. Pale pink and gold sunrise light, light haze. Elegant, realistic, graceful motion.

Shot 1: Extreme close-up of a pastry chef’s tweezers placing a single mint leaf on a glossy chocolate dessert; shallow depth of field, static camera.
Cut to
Shot 2: The camera pulls back to a wide shot of the same chef in a busy restaurant kitchen, cooks moving behind her, steam rising. Warm tungsten light, polished commercial style.

Tips for better Seedance results

• check_circleRepeat key identity details (hair, clothing, age) in every shot to avoid characters changing between cuts.
• check_circleGive each shot a different framing so cuts feel intentional: wide, then close-up, then medium.
• check_circleAssign one camera move per shot; mixing moves inside a shot reduces stability.
• check_circlePut global style and lighting at the top so you do not have to repeat them for every shot.
• check_circleUse image-to-video with a reference image when consistency matters more than variety.
• check_circlePick the aspect ratio for your destination first; reframing later often crops important action.

Mistakes to avoid

• check_circleWriting several shots as one paragraph without clear separators.
• check_circlePacking too many shots into a short clip, which leaves each one only a moment on screen.
• check_circleChanging style between shots by accident (for example “anime” in shot 1 and “photorealistic” in shot 2).
• check_circleRelying on long on-screen text, which video models rarely render well.
• check_circleUsing names of real people or copyrighted characters, which most platforms block.

How the generator helps

The tool above sends your idea, plus any options you chose, to an AI model with instructions tailored to Seedance: split distinct moments into numbered shots, repeat the character description in every shot, give each shot its own framing and camera move, and mark every cut. Options left on “Auto” are chosen by the AI; options you set are respected exactly. If you only describe one moment, you get a single continuous shot instead.

If the AI service is busy, or you have reached the free limit, the page tells you plainly and does not invent a result. Generate a few versions, paste the best one into Seedance, and refine from there.

Related tools

Want each shot as a separate structured object? Switch the output to JSON, or design custom fields in the JSON Prompt Generator. To break a longer story into scenes first, use the Video Prompt Generator. To learn from a clip you admire, upload it to Video to Prompt and rewrite the result as Seedance shots.

Frequently asked questions