All articles

Article

How to Write Better Seedance 2.5 Video Prompts

Seedance 2.5 is ByteDance's longest multimodal video model on HiArt: up to 30 seconds, native synced audio, and omni-reference for up to 50 media files. Results still depend on how you write the prompt — especially how you assign roles when many @ references share one generation.
This guide covers the syntax and patterns that work for longer, reference-rich clips. For a product overview, start with What Is Seedance 2.5? or the model page.

What you can upload

Plan assets before writing — Seedance 2.5 accepts far more files than Seedance 2.0, so unused uploads waste credits and confuse the model.

Input limits

Input
Limit
Format
Max size
Images
≤ 30
jpeg, png, webp, bmp, tiff, gif
30 MB each
Videos
≤ 10
mp4, mov
50 MB each
Audio
≤ 10
mp3, wav
15 MB each
Text
Natural language prompt
Total
≤ 50 files

Output specs

  • Duration: 4–30 seconds (you choose)
  • Native synced dialogue, SFX, and background music with lip sync
  • Resolution: 480p or 720p

Good to know

  • Realistic human faces in uploads are blocked for compliance — use stylized or non-realistic references instead.
  • Mention every uploaded file in the prompt; orphan references hurt consistency.
  • For 20–30s scripts, write timed beats — do not dump a single untimed paragraph.
  • Need 1080p under 15s instead? Switch to Seedance 2.0 after locking the prompt on 2.5 or Fast.

The @ reference system

Every uploaded asset gets a slot — @Image1, @Video2, @Audio1, and so on. The model does not infer what each file is for. You must say it explicitly in the prompt.
@Image1    @Image2    @Image3  … up to @Image30
@Video1    @Video2   … up to @Video10
@Audio1    @Audio2   … up to @Audio10

Common reference roles

Purpose
Example
First frame
@Image1 as the first frame
Last frame
@Image2 as the last frame
Character look
@Image1's character as the subject
Identity lock across shots
keep the character consistent with @Image1 @Image2 @Image3
Scene / background
scene references @Image8
Camera movement
reference @Video1's camera movement
Action / choreography
reference @Video1's action choreography
VFX / transitions
completely reference @Video1's effects and transitions
Rhythm / tempo
video rhythm references @Video1
Background music
BGM references @Audio1
Voice / dialogue tone
voice tone references @Audio2
Sound effects
sound effects reference @Video3's audio
Stack roles when the shot needs them — especially useful with 2.5's larger reference budget:
@Image1's character as the subject, wardrobe from @Image2, scene references
@Image3 @Image4, camera and pacing from @Video1, BGM references @Audio1,
dialogue tone references @Audio2

How to structure a prompt

Strong prompts usually follow this order:
[Subject] + [Scene] + [Action/Motion] + [Camera] +
[Timing] + [Effects/Transitions] + [Audio] + [Style/Mood]
For clips longer than about 10 seconds — and especially 20–30 second Seedance 2.5 renders — break the action into timed beats:
0–5s: opening hook, establish subject and location
5–12s: develop action / product benefit
12–20s: climax, key reveal, or dialogue turn
20–30s: resolution, brand end card, music settle

Camera language cheat sheet

Seedance responds well to standard film terminology. Mix movement + shot size for finer control.

Movements

Term
What it does
Push in / Slow push
Camera moves toward subject
Pull back
Camera moves away from subject
Pan left/right
Horizontal rotation
Tilt up/down
Vertical rotation
Track / Follow shot
Camera follows subject
Orbit / Revolve
Camera circles subject
One-take / Oner
Single continuous shot, no cuts

Shot sizes

Term
Frame
Extreme close-up
Eyes, mouth, or small detail
Close-up
Face fills frame
Medium shot
Waist up
Full shot
Entire body
Wide / Establishing
Full environment

Examples that work on Seedance 2.5

Keep a character consistent across many refs

Use several look images, then pace a longer scene:
Keep the character consistent with @Image1 @Image2 @Image3. Wardrobe from
@Image4. Scene references @Image5.
0–6s: The character walks into a warm apartment hallway, medium shot,
natural dialogue under breath.
6–14s: Close-up — they pause at the door, exhale, soft smile.
14–22s: Door opens; child and pet rush in. Interior references @Image6.
22–30s: Wide shot of the living room, warm practical lights, music settles.
BGM references @Audio1. Voice tone references @Audio2.

Multi-ref music-driven cut

@Image1 through @Image8 are keyframe looks for the same performer.
Match the cut rhythm of @Video1 and the beat of @Audio1.
0–8s: Close-ups and handheld energy, lip sync to @Audio1.
8–18s: Wider dance moves; completely reference @Video2's choreography.
18–30s: Orbit around the performer, neon city lights, final freeze on logo.
More dynamic motion; dreamlike grade; beat-synced transitions.

Copy camera work, then extend the story

Reference @Image1's character in @Image2's elevator.
Completely reference @Video1's camera movements and facial expressions
for the first half.
0–8s: Hitchcock zoom during the fear beat, then orbit the cabin.
8–16s: Doors open; follow shot into a lobby that references @Image3.
16–24s: Character walks to glass exit; rain streaks on windows.
24–30s: Exterior night street, pull back to wide, music drops to silence.
Ambient SFX; tense piano BGM.

30s product story with voiceover

Hero product is @Image1. Lifestyle context from @Image2 @Image3.
Editing energy references @Video1. Voiceover cadence references @Audio1.
0–5s: Product rotates into frame, macro on texture and logo.
5–12s: Angle transitions — front, side, detail — with scanning highlight.
12–20s: Product in real usage; hands and environment stay clean.
20–26s: Benefit line as natural dialogue / VO synced to picture.
26–30s: Hero end frame, brand tagline, music resolves.
Sound: BGM references @Audio2; add subtle product interaction SFX.

Style and audio modifiers

Append short phrases at the end of your prompt to steer look and sound:
  • Visual: Cinematic quality, film grain, shallow depth of field · 2.35:1 widescreen, 24fps · Anime style / Photorealistic
  • Mood: Tense and suspenseful · Warm and healing · Epic and grand
  • Audio: Background music: grand and majestic · Beat-synced transitions · Lip-synced dialogue · Voice tone reference @Audio1

A quick workflow for longer clips

  1. Decide the video type — ad, drama, MV, product story, educational.
  2. Gather only the references you will cite; map each to an @ slot.
  3. Write timed beats for the full duration (especially above 15s).
  4. Assign a role to every @ reference — look, camera, BGM, VO, etc.
  5. Generate at 480p/720p on Seedance 2.5; tighten ambiguous lines next pass.
  6. If you later need 1080p under 15s, port the locked prompt to Seedance 2.0.

Mistakes to avoid

  1. Vague references — "reference @Video1" without saying camera, action, or rhythm.
  2. Unused uploads — twenty images attached but five mentioned in the prompt.
  3. Untimed 30s scripts — a wall of text with no beat markers rarely lands.
  4. Conflicting camera notes — static camera and orbit in the same segment.
  5. Overpacked early seconds — three scene changes in the first 4 seconds.
  6. No audio direction — on 2.5, sound and lip sync are major quality levers.
  7. Duration mismatch — a 30-beat outline with 10-second output selected.

Copy-paste templates

30s product ad

Reference @Video1's editing style and camera transitions. Replace @Video1's
product with @Image1 as the hero product. Lifestyle frames @Image2 @Image3.
0–5s: Product enters with dynamic rotation, close-up on texture and logo.
5–12s: Angle transitions — front, side, back — with highlight scanning light.
12–20s: Product in lifestyle usage context with natural motion.
20–26s: Short synced VO describing the key benefit.
26–30s: Hero shot, brand tagline, music builds to resolution.
Sound: Reference @Audio1 for BGM. Add product interaction SFX.

24s short drama

Characters consistent with @Image1 and @Image2. Interior references @Image3.
0–6s: Close-up on reddened eyes, finger pointing, tears. Dialogue (A):
"What exactly are you trying to take from me?"
6–14s: Other character trembles, holds evidence, steps forward. Dialogue (B):
"I'm not deceiving you! This is what he entrusted to me!"
14–20s: Evidence revealed. A freezes — anger shifts to shock.
20–24s: Wide shot; rain on the window; music softens to a single piano note.
Sound: Urgent piano early, static interference, sobbing, button click.

20s dance / lip-sync clip

Have the character in @Image1 replicate the dance moves from @Video1 and
lip-sync to @Audio1. Generate a 20-second video.
0–8s: Tight framing, sharp beat hits, clear mouth shapes.
8–16s: Wider full-body choreography, smooth motion, no stuttering.
16–20s: Orbit finish, freeze on pose, music ring-out.
Movements smooth; neon stage lighting; vertical 9:16.

Ready to try it? Open a video node on HiArt canvas, select Seedance 2.5, upload only the references you will cite, and paste a timed template from this guide. Start with one character look, one clear action arc, and one audio reference — then layer more @ slots as the longer timeline needs them. Compare duration vs resolution tradeoffs in Seedance 2.5 vs 2.0.