Features
Customers
Resources
BACK

Seedance 2.0 Prompt Guide 2026

Seedance 2.0 Prompt Guide 2026

Seedance 2.0 Prompt Guide 2026

Feb 21, 2026

Feb 21, 2026

A significant change in AI video production appears in Seedance 2.0. This model takes text, images, videos, and audio as inputs instead of just text prompts or a single reference image, letting you control every part of your creation like a real filmmaker.

Its reference capability is its most notable feature; you can use an image to set the visual style, a video to specify motion and camera work, audio to drive the rhythm, and text to direct the story. As a result, generative video can now be controlled to a degree that was previously unattainable.

Seedance 2.0 lets you upload up to 12 files (9 images + 3 videos max 15s total + 3 audio max 15s total) and reference them with @ tags to control exactly what the AI generates. This guide shows you how to write prompts that actually work.

Access Seedance on Atlabs AI.

What You Can Upload

  • 9 images - Characters, environments, style references

  • 3 videos (15s total) - Camera movements, action choreography

  • 3 audio files (15s total) - Music, dialogue, sound effects

Reference them in prompts using @Image1, @Video1, @Audio1

How @ References Work

Simple example:

Upload: Character photo + landscape photo + music
Prompt: "Character from @Image1 walks through @Image2 landscape.
Camera tracking shot. Beat-synced to @Audio1."

What to reference from each file type:

File Type

Use For

@Image

Character faces, environments, style, first/last frames

@Video

Camera movements, action choreography, pacing

@Audio

Music timing, dialogue lip-sync, ambient sound

Pro tip: Start with 2-3 files, not 12. More ≠ better.

Access Seedance on Atlabs AI.

Template 1: Timeline Structure (For Story Videos)

Use this for trailers, music videos, narrative content.

【Basic Settings】
Duration: 15 seconds
Style: [Your visual style - e.g., "Cinematic epic fantasy, Unreal Engine quality"]
【Timeline】
0-5s (Setup):
Visual: [Opening scene description]
Camera: [Movement type - dolly-in / aerial / tracking]
Details: [Physical elements - wind, cloth physics, particles]
5-10s (Conflict):
Visual: [Action/tension description]
Lighting: [Light effects - glows, reflections, shadows]
Action: [Specific movements and impacts]
10-15s (Climax):
Visual: [Peak moment description]
Environment: [Scene changes - weather, destruction]
Ending: [Final shot - freeze frame / zoom out / fade]
【Audio】
SFX: [Environmental sounds]
Dialogue: [Any spoken lines with emotion]
Music: [Mood and style]

Real Example:

Style: Lord of the Rings epic, cinematic dark lighting
Reference: @Image1-3 for three-panel storyboard
【Timeline】
0-5s: Camera descends from mountains to reveal Fellowship walking through snow. Cloaks flapping, breath visible. Reference @Video1 for crane movement.
5-10s: Cut to Moria. Gandalf fights goblins with glowing staff (@Image4). Aragorn and Legolas coordinate attacks. Sparks flying, troll roaring. Reference combat timing from @Video1.
10-15s: Gandalf faces Balrog at bridge. Staff glows intensely. "You shall not pass!" Bridge collapses. Freeze on Gandalf falling. Zoom out to darkness.
SFX: Snow crunch, wind howling, weapon clashes, bridge collapse thunder
Dialogue: "You shall not pass!" (firm, solemn, reverb)
Music: Reference @Audio1 for orchestral epic timing

Access Seedance on Atlabs AI.

Template 2: Nine-Panel Structure (For Action/Character Content)

Use this for fight scenes, character showcases, game trailers.

Step 1 - Environment:
[Location + atmosphere - e.g., "Ancient temple ruins, stormy sky, lightning flashes"]
Step 2 - Character:
[Appearance + equipment + emotion - e.g., "Knight in silver armor, glowing sword, determined expression"]

Upload character as @Image1
Step 3 - Nine Panels (5 seconds each):
Line 1 (0-5s Setup):
Panel 1: [Wide establishing shot]
Panel 2: [Push to character]
Panel 3: [Enemy appears]

Line 2 (5-10s Action):
Panel 4: [Attack begins]
Panel 5: [Impact/effects]
Panel 6: [Enemy defeated]

Line 3 (10-15s Resolution):
Panel 7: [Aftermath wide shot]
Panel 8: [Character detail/cost]
Panel 9: [Future tease/ending]

Step 4 - Style:
[Visual keywords - Cinematic lighting, epic scale, etc.]

Example 1:

Step 1 - Environment:
Crumbling gothic cathedral at twilight. Storm clouds, purple sky, lightning. Magical energy crackling. Ancient runes glowing blue on stones. Reference @Image1 for environment.

Step 2 - Character:

Paladin in silver plate armor with gold trim. Two-handed glowing greatsword. White cape. Lion crest on chest. Helmet with cross visor. Righteous stance. Reference @Image2 for character (must stay identical throughout).

Step 3 - Nine Panels:

Line 1 (0-5s):
Panel 1: Crane shot descending through clouds to cathedral. Lightning flashes. Reference @Video1 for camera.
Panel 2: Paladin walks through ruins. Cape billowing. Medium to close-up push.
Panel 3: 20+ cultists with red eyes emerge from shadows. Surround paladin. Reference @Image3 for cultist design.

Line 2 (5-10s):

Panel 4: Paladin raises sword. Holy runes ignite gold-white. Battle cry with reverb. Charges forward. Tracking shot. Reference @Video1 for combat timing.
Panel 5: Sword swing creates light trail. Collides with dark magic. Massive shockwave. Slow-mo 1 second. Orbit camera 180° around impact.
Panel 6: Cultists disintegrate to ash. Pillars shatter. Dust cloud. Paladin silhouette in gold light.

Line 3 (10-15s):

Panel 7: Crane up. Smoke clears. Battlefield ruins revealed. Storm breaking. Sunlight beams (volumetric rays).
Panel 8: Close-up sword still glowing. Tilt up to helmet. Battle-worn. Heavy breathing.
Panel 9: Paladin turns to face distant mountains. Dark storm there. Wide shot silhouette. Freeze. Fade black.

Step 4 - Style:

Volumetric god rays, cinematic lighting, epic scale, dynamic magic effects, medieval dark fantasy, gritty textures, heroic composition. --ar 16:9
Audio: Thunder, wind, stone grinding, magic crackling, sword impacts. Battle cry. Reference @Audio1 for epic orchestral score.

Access Seedance on Atlabs AI.

Example 2 :

Cinematic style: Dark, moody kitchen lit by a single overhead pendant light. Handheld camera, shallow depth of field. Color grade: desaturated with cold blue-green tones. Two actors, late 20s-early 30s. Tense, intimate, thriller tone.
SHOT 1 (0:00–0:03) — CLOSE-UP, Sarah's eyes Tight close-up of Sarah's face from the nose up. Her eyes widen slightly with fear. She says flatly: "You went through my phone." Kitchen background softly blurred. Subtle camera drift to the right.
SHOT 2 (0:03–0:05) — OVER-THE-SHOULDER, Mark Camera behind Sarah's shoulder, focused on Mark standing across the kitchen island. His jaw is tight. He holds up her phone and says coldly: "You gave me a reason to." He sets the phone down with a deliberate tap on the counter.
SHOT 3 (0:05–0:08) — MEDIUM TWO-SHOT, side profile Both characters framed in profile facing each other across the counter, pendant light hanging between them. Sarah swallows hard and asks: "What did you find?" Mark responds: "Nothing. That's what scares me. You deleted everything." Slight push-in on the camera.
SHOT 4 (0:08–0:12) — CLOSE-UP, Sarah's mouth and chin Extreme close-up of Sarah's trembling lips. A tear catches the light on her chin. Her voice cracks: "I was protecting you." Mark's voice off-screen: "From what?" She whispers: "From what I almost did."
SHOT 5 (0:12–0:15) — WIDE SHOT, kitchen doorway Static wide shot from behind Sarah. Mark stares at her for one beat, then turns and walks through the dark doorway, disappearing into shadow. The phone sits on the counter glowing faintly. Sarah stands alone under the light. Silence. Cut to black.

Quick Prompting Tips

Motion Intensity

Must specify speed - AI can't infer from images.
❌ "Car drives by"
✅ "Race car speeds past at 120 mph, tires screeching, motion blur"

Use these words: quickly, slowly, dramatically, vigorously, frantically, powerfully

Camera Movements

Movement

When to Use

Example

Dolly

Cinematic approach/reveal

"Slow dolly-in on character's face"

Pan

Show adjacent space

"Slow pan right revealing cityscape"

Tracking

Follow action

"Tracking shot alongside running athlete"

Crane

Dramatic vertical

"Crane down from sky to ground level"

Handheld

Raw/documentary feel

"Shaky handheld follow, natural movement"

Gimbal

Smooth professional

"Stabilized gimbal tracking, polished"

For compound moves: "Start: slow dolly-in for 3 seconds. Then: gentle pan right for final 2 seconds"

Character Consistency (Stop Face Morphing)

Do this:

  • Upload 2K+ resolution character images

  • Use @Image1 as your character reference in every prompt

  • Add: "Character from @Image1 must stay identical. Same face, hair, outfit."

  • Keep design simple (fewer accessories = more stability)

Don't:

  • Use low-res images

  • Upload multiple different character angles without specifying main reference

  • Leave consistency to chance

Multi-Shot Sequences

Shot 1 (0-5s): Wide - Knight enters forest, dolly-in through trees
Shot 2 (5-10s): Medium - Knight draws sword, orbit camera
Shot 3 (10-15s): Close-up - Knight's eyes, slow push-in
Transition: Shot 1→2 quick cut | Shot 2→3 match cut on sword
Character: @Image1 appearance locked throughout

Common Issues & Quick Fixes

Problem

Fix

Face morphing

Use 2K+ image. Add "@Image1 character stays identical"

Jittery motion

Specify speed: "slow dolly-in" not just "dolly"

Warped hands

Simple poses only. High-res reference images

Scene discontinuity

Lock environment to one @Image reference

AI ignores prompt

Keep 30-100 words. Use 2-4 references max

When generation fails:

  1. Framing wrong? → Change camera (shot size + movement only)

  2. Motion wrong? → Swap handheld ↔ gimbal, add speed

  3. Style drifts? → Use single style anchor, remove extra adjectives

  4. Character mutates? → Simplify description, add "no extra people"

Where to Use Seedance 2.0

Access Seedance on Atlabs AI.

Best for: Social media (TikTok, Reels, Shorts), product demos, game trailers, music videos, tutorials

FAQ

Q: How many files should I upload?
A: 2-3 carefully chosen assets work best. Quality > quantity.

Q: What resolution for images?
A: Minimum 1080p, ideally 2K-4K.

Q: Can I extend videos beyond 15s?
A: Yes, generate 15s then use as reference to extend.

Q: Does it support negative prompts?
A: No. Use positive constraints instead.

Q: Seedance 1.5 vs 2.0?
A: 2.0 adds multimodal input (video+audio), @ references, native audio, 3× longer.

Start Creating

Access Seedance on Atlabs AI .

Ready? Copy a template, upload your assets, and create professional AI videos in minutes.

Ready to tell your story?

Ready to tell your story?

Ready to tell your story?