Seedance Pro Builder
Cinematic Seedance 2.5 video prompts, up to 30-second shots · 32 fields · 5 outputs
Released 31 July 2026. Up to 30 seconds in a single pass, up to 4K output and up to 50 multimodal references. Default model in Dreamina/Jimeng; API since August 2026.
✨ Enhance with AI
Rewrite the main prompt into 3 polished variations. Pro feature.
01Main Seedance Video Prompt
Medium shot, low angle, slow dolly in, real time.
Young woman, streetwear outfit, walking toward camera, confident expression, wind in the hair.
Setting: neon city street, golden hour, clear sky.
Rim light, shallow depth of field bokeh, cinematic atmosphere.
Color grade: teal and orange. Style: photorealistic, moderate motion.
Audio: cinematic orchestral, ambient city noise, beat synced cuts.
4K, high detail. Avoid: text or watermarks.
02Camera Movement Brief
Camera movement: slow dolly in.
Angle: low angle. Framing: medium shot.
Speed: real time, moderate motion.
Color grade: teal and orange. Keep the camera path smooth and continuous for the full 10 seconds.
03Reference File Brief
1. No character reference selected. Choose one in section 06 to lock the face and outfit across shots.
2. Optional: add a style reference to keep the look identical across a series.
3. Prompt: @image1 as main subject, young woman walking toward camera, neon city street.
4. Re-use the same references and seed for every shot in the sequence.
04Audio Sync Brief
Music: cinematic orchestral.
Cuts and camera moves land on the beat.
Sound effects: ambient city noise.
Voice: no voice.
Mix: beat synced cuts, mood cinematic.
05Multi-Shot Sequence
Shot 1 — Establishing: extreme wide shot, neon city street, slow dolly in.
Shot 2 — Introduce: medium shot, young woman walking toward camera, slow dolly in.
Shot 3 — Emotion: close up, confident expression, slow dolly in.
Shot 4 — Detail: macro shot, wind in the hair, slow dolly in.
Shot 5 — Closing: wide shot, young woman in neon city street, cinematic ending, slow dolly in.
Style lock: photorealistic, teal and orange, rim light.
Keep the same subject in every shot.
📘 Pro Guide
Model status (checked Nov 2026)
- Seedance 2.5Released 31 July 2026. Up to 30 seconds in a single pass, up to 4K output and up to 50 multimodal references. Default model in Dreamina/Jimeng; API since August 2026.
- Seedance 2.0Native audio-video model from February 2026, upgraded to native 4K with 10-bit colour in June 2026. Seedance 2.0 Fast trades some quality for speed.
Prompting tips
- One subject, one actionName the subject first, then a single clear action. Extra actions in one shot are where faces and hands start to morph.
- Camera before styleState the camera move, angle and shot size before the look. Seedance follows camera language closely.
- ReferencesTag uploads as @image1, @image2 or @video1 and say what each one controls: character, style, colour or camera motion.
- Long clipsFor 20 to 30 second clips on 2.5, describe the beats in order (opening, middle, ending) so the motion has somewhere to go.
- AudioDescribe music, sound effects and voice in the prompt. For dialogue, write the exact line in quotes and choose lip synced dialogue.
- NegativesKeep negatives short: no text or watermarks, no morphing faces, no flicker. Long negative lists dilute the prompt.