Protect the reference first
Name the parts of your subject that can't change: logo, shape, outfit, face, material, color, or pose. This gives Seedance 2 a stable anchor before it touches motion.
Seedance 2 · Multimodal reference video generation
Seedance 2 turns any reference into controlled video.
Seedance 2 reads text, images, video clips, and audio in a single prompt, then keeps your product, character, motion style, and sound locked to the references you gave it. Use text to video to start from an idea, image to video to animate a still, or multi-reference generation when a shot needs several anchors at once.
Free to start · no credit card needed for a draft
- Up to 9 images
- 3 videos + 3 audio files
- 4s-15s clips
Seedance 2
Multi-reference, text, image, and native-audio video generation
240/4000
Reference images
0/9 · JPG/PNG/WebPAudio
See the output
Four ways to start a shot, plus the controls that keep it usable
Preview Seedance 2 as a production path: start from text, still frames or multiple references, then check whether the output keeps motion, subject and camera direction under control.
0:15
Porcelain runway illusion
0:15Photoreal Hollywood haute-couture fantasy on a mirror salt lake under oppressive storm clouds. A model walks in a liquid blue-and-white porcelain dress; one snap shatters it into ink-wash swallows as the lake dissolves into a black-white 3D fluid ink vortex.
A practical workflow for Seedance 2 reference videos
Describe one clear action
Give the clip a readable action: open, pour, walk, rotate, scroll, reveal, lift, compare, transform. Seedance 2 performs better when the motion is specific rather than implied.
Add the environment after the action
Once the subject and action are locked, layer in light, background, camera move, pacing, and optional sound. That keeps the video useful instead of merely decorative.
Seedance 2 generation parameters
480p-4K
Resolution
Draft fast at 480p, then finish at 1080p or 4K when a shot is ready to ship
9+3+3
References
Combine up to 9 images, 3 videos, and 3 audio files in one Seedance 2 prompt
<15s
Reference media
Reference videos and audio files share a 15-second combined limit
4-15s
Duration
Set clip length directly in the generation panel
Built like a film crew, not a filter.
Seedance 2 plans picture, motion, camera, and sound in one pass. Reach for it when a clip needs synchronized audio, stable action, consistent subjects, and camera movement that actually follows the prompt.
Audio0:15
Audio
Synchronized audio, one pass
Describe dialogue, effects, ambience or music cues in the same prompt. Seedance 2 generates the sound with the picture instead of treating audio as a separate step.
Motion0:15
Motion
Stable motion through action
Guide characters, products and environments to keep timing, weight and momentum clear as the shot moves from one action to the next.
Subject0:15
Subject
Subject consistency across the clip
Keep faces, outfits, product shapes, materials and scene logic connected so the result reads as one take, not several disconnected generations.
Camera0:15
Camera
Prompt-guided camera movement
Ask for one main camera path such as a push-in, orbit, tracking shot or reveal, then let Seedance 2 carry that direction through the whole clip.
Seedance 2 image to video prompts that keep the subject intact
Every Seedance 2 prompt below starts by protecting what can't drift, then layers in one clear action, a camera move, and an environment cue — built for product, creator, ecommerce, and character clips.
Prompt 01
Ceramic mug image to video
Animate the provided product photo of a speckled ceramic mug without changing its shape, glaze pattern, handle, or color. Steam rises from the cup while the camera makes a gentle push-in across a wooden breakfast table. Keep the mug fully recognizable, add warm window light, a soft spoon clink, cozy ecommerce lifestyle mood.
Prompt 02
Denim jacket fit demo
Use the reference image of the denim jacket as the exact product. A model turns from front to side, lifts one arm to show the fit, then relaxes into a natural stance. Preserve wash color, seams, buttons, pocket placement, and silhouette. Clean studio daylight, smooth half-orbit camera, practical fashion product video.
Prompt 03
Mascot sticker walk cycle
Take the supplied mascot illustration and turn it into a short walking loop through a bright stationery shop. Keep the authorized mascot design, proportions, colors, backpack, and outline style. The character waves once, passes shelves of notebooks, and exits frame right. Gentle follow camera, cheerful but not exaggerated motion, no style drift.
Prompt 04
Watch packaging unbox
Start from the reference watch box image and keep the box dimensions, color, logo position, and material finish unchanged. A hand opens the lid slowly, revealing the watch cushion inside while the camera moves from top-down to a low three-quarter angle. Soft studio sound, premium unboxing feel, stable product identity throughout.
Prompt 05
Fitness app phone tilt
Use the uploaded phone screenshot as the fixed screen design. A hand lifts the phone from a gym bench, tilts it toward the camera, then taps one workout card. Preserve the app layout, colors, and visible interface hierarchy without inventing new screens. Gym ambience, rubber mat texture, clear creator demo motion.
Prompt 06
Bakery window pastry reveal
Animate the reference pastry photo while keeping the exact pastry shape, topping pattern, plate, and table setting. The camera slides along a bakery window as sunlight moves across the glaze and a hand gently rotates the plate. Add quiet street ambience, a soft plate sound, appetizing realistic motion, no extra decorations.
Seedance 2 compared with the other video models
Same shot brief, four different jobs. Seedance 2 sits in the standard multimodal lane, with broader references, higher resolution options, and more control than the Fast tier — the pick for teams choosing between Seedance 2 and Veo for reference-driven work.
| Feature | Veo 3.1 | Veo 3.1 Lite | Seedance 2 | Seedance 2 Fast |
|---|---|---|---|---|
| Best for | Premium cinematic clips | Fast Veo-style drafts | Reference-driven motion | High-volume variations |
| Finish | Highest finish | Lighter pass | Flexible | Quick draft |
| Audio | Native sync audio | Sync audio | Optional audio | Simple audio |
| Speed | Standard | Faster | Standard | Fastest |
| Resolution | Up to 1080p | Up to 1080p | 480p-4K | 480p–720p |
| Top use | Hero ads, brand films | Social hooks, iteration | Ecommerce, character | Batch ad testing |
FAQ
FAQ
Pick the right video model for the shot
Seedance 2 Fast
Seedance 2 Fast keeps the same practical input modes as the standard model — multi-reference, text to video, first/last-frame image to video — but tunes them for quicker iteration. Use it to test hooks, products, UGC-style moments, and reference combinations before promoting the winner to a higher-finish pass.
Compare this modelVeo 3.1
Reach for this model when the clip has to sell a moment: a product reveal with real camera intention, a founder ad with natural motion, a scene that carries atmosphere, sound, and continuity from the first frame to the last. Built on Google DeepMind's Veo, it follows detailed direction on lighting, movement, and audio in one pass.
Compare this modelSeedance 2 · Multimodal reference video generation