Seedance 1.0 Pro: how to write prompts the model actually understands

What 1.0 Pro can and cannot do

1.0 Pro is the «short, fast, reliable» video generator. Duration is locked at 5 or 10 seconds — no 7 or 12. Resolution is 720p or 1080p, FPS is always 24. Inputs are text or one image for I2V. Pro is faster than 2.0 and more stable on simple scenes.

First-generation limits apply across the 1.0 line: no @-references (only a single unprefixed image), no Consistency Control between runs, no sound control, no video extension, no second-by-second storyboarding. Negative prompts do not work at all — the model ignores them or breaks on them.

The strong suit is multi-shot with camera switching: «Cut to», «Camera cut to», «Camera switching» inside a 10-second clip work more reliably than in 1.0 Lite.

• Duration locked: 5 or 10 seconds
• Resolution 720p / 1080p, 24fps
• Input: text or 1 image (I2V)
• Multi-shot with camera switching works
• Negative prompts do NOT work

Prompt structure

Canonical formula: `[Subject] + [Action] + [Scene] + [Camera] + [Style]`. The core is subject, action, scene. Camera and style amplify but do not replace the core.

For Text-to-Video: «Subject + Motion + Scene + Camera + Style». For Image-to-Video the key is dynamics, not static description: «Subject + Motion, Background + Motion, Camera + Motion». An I2V prompt that describes what is already visible in the reference leaves the model guessing about what should move.

Always specify **fixed camera** for a locked-off shot or **non-fixed camera** for any movement. Without it the model picks unpredictably. Base parameters — resolution, aspect ratio, duration — go at the end: «Non-fixed camera, 720p, 9:16, 5s».

Camera switching and multi-shot

The main feature 1.0 Pro has over Lite is the ability to switch shots inside a single video. Key phrases: «Cut to», «Camera cut to», «Camera switching». In a 10-second clip you can fit 2–3 shots.

Documentation example: «Panoramic shooting, the model approaches with a smile. Camera switching, close-up of the lower body, the straight design of the pants and the drape of the fabric are highlighted while walking.» One prompt describes a wide shot → switch → tight shot of the lower half.

On a scene cut via «Cut to» — describe the new scene in words after the switch, do not leave «Cut to.» hanging. There has to be a logical link between shots, otherwise the model produces a hard discontinuity.

Intensity adverbs

Without explicit intensity 1.0 Pro renders «medium» motion — slow, smooth, no energy. To get dynamics you need adverbs: fast, intense, large, high frequency, strong, crazy, quickly.

Examples: «car quickly passing by» instead of «car passing by» actually delivers a fast drive-by. «Wings flapping wildly» instead of «wings flapping» — wings working at full power. «Crazy fast camera dolly-in» instead of «camera dolly-in» — a sharp push-in.

Adverbs work both on camera moves and on character actions. This is the cheapest way to wake up a sluggish clip — add 2–3 adverbs at key moments.

Common mistakes

1. 1. Negative prompts in 1.0 Pro

   «No watermark», «no cartoon», «without blur» — 1.0 Pro either ignores these or breaks on them. Rule: state what to show, not what to hide. Instead of «not cartoonish» → «photorealistic, film grain». Instead of «no text» → describe scenes without surfaces for text.

2. 2. @-reference syntax

   @image1, @video1, @audio1 are all from 2.0. 1.0 Pro doesn't have them: a single image input without a prefix for I2V, and that's it. A prompt written with @-syntax leaks the tokens into the text as garbage and quality drops.

3. 3. Asking for 7, 12, or 15 seconds

   Duration is locked to 5 or 10 seconds. No in-between options. «Make an 8-second clip» either gets cropped to 5 or stretched to 10 with weird pacing. If you need 15, that's 2.0 or New.

4. 4. I2V prompt describes what is already in the photo

   If the reference shows a car and the prompt says «red car on a road», the model has no idea what should move. I2V formula: «Subject + Motion, Background + Motion, Camera + Motion». Describe the dynamics — what accelerates, what rotates, where the light flies.

5. 5. Second-style timestamps like «0-3s / 4-8s»

   Timestamp storyboarding arrived in 2.0. 1.0 Pro does not parse «0-3s» or «4-8s» as structure. For multi-shot use «Cut to» and «Camera switching» with an explicit description of each shot.

Before / after examples

beautiful model in a dress walking

Full-body shot of a young woman in a long red silk dress walking down a marble corridor. She turns slightly toward the camera with a confident smile. Panoramic shooting, slow tracking shot following her movement. Camera switching, close-up of the dress fabric drape and her hand brushing the silk. Soft natural daylight from tall windows on the right, golden hour warm tones. Non-fixed camera, 1080p, 9:16, 10s.

don't make it cartoonish, no text and no watermarks

Cinematic close-up of a steaming espresso cup on a dark wooden table in a softly lit café. Steam rises slowly, light catches the foam surface. Camera slowly orbits around the cup, ending on a side angle that reveals the latte art. Warm tungsten lighting, shallow depth of field, film grain. Non-fixed camera, 1080p, 16:9, 5s.

animate my car photo

The sports car quickly accelerates from a standstill on the wet asphalt road. Tires spin briefly, water spray rises from behind. Camera follows from a low rear-tracking angle, then switches to a side shot showing the full car body in motion. Headlights glow in the dim evening light, reflections on the wet road. Non-fixed camera, 1080p, 16:9, 5s.