Seedance 2 Video Generator
seedance2 is a multi-modal video model with reference-anything control, precise motion & camera replication, video extension/editing, and built-in audio generation.
Tip: in seedance2, specify what each reference controls (style/subject/motion/camera/audio) and add audio cues for tighter sync.
seedance2 prompt examples
Curated seedance2 clips with real prompts from the official guide.
Laundry realism
Prompt: seedance2 A girl elegantly hangs laundry. After finishing, she takes another piece from the bucket and shakes it hard.
Street chase action
Prompt: seedance2 The camera follows a man in black running fast, with a crowd chasing behind. The camera shifts to a side tracking shot; he panics, knocks over a fruit stall, gets up, and keeps running, with the chaotic sound of the crowd.
Family reunion continuity
Prompt: seedance2 The man @image1 walks down the corridor after work, tired, slows down, and stops at his door. Close-up of his face: he takes a deep breath, steadies his emotions, and relaxes. Close-up of him finding the key, inserting it into the lock, and entering. His little daughter and a pet dog run over to greet and hug him; the interior is warm, with natural dialogue throughout.
Luxury bag product shot
Prompt: seedance2 Commercial showcase of the bag in @image2. The side view references @image1; the surface material references @image3. Show every detail of the bag; grand, cinematic background audio.
Elevator suspense camera
Prompt: seedance2 Reference the man in @image1; he is in the elevator from @image2. Fully match all camera moves and facial expressions from @video1. Use a Hitchcock zoom when he is terrified, then several orbit shots to show the elevator interior. The doors open; follow him out. The exterior scene references @image3. The man looks around; reference @video1 for robotic-arm multi-angle follow of his gaze.
Rotating tablet product shot
Prompt: seedance2 Use the tablet in @image1 as the main subject. Camera motion references @video1, pushing in to a close-up of the screen. After rotating, the tablet flips to show the full body; the data flow on the screen keeps changing. The surroundings gradually become a sci-fi data space.
Sci‑fi POV template
Prompt: seedance2 Replace the person in @video1 with @image1; @image1 is the first frame. The person puts on virtual sci-fi glasses. Reference @video1 for camera moves and close orbit shots, shifting from third-person to POV. Travel through the AI glasses to the deep blue universe in @image2; several ships fly into the distance. The camera follows the ships into the pixel world in @image3, flying low over a pixel forest where tree growth forms appear. Then the view tilts up and rapidly travels to the light-green textured planet in @image4, sweeping over its surface.
Particle title opener
Prompt: seedance2 Start on a black screen. Reference @video1 for particle effects and material; golden gilded sand particles drift from left to right to cover the frame. Reference @video1 for the particle blow-away effect; the text in @image1 gradually appears at the center.
Comic panel animation
Prompt: seedance2 Adapt @image1 into a comic sequence from left to right and top to bottom. Keep the dialogue consistent with the image. Add special sound effects for panel transitions and key moments. Overall style is humorous; the presentation references @video1.
Seasonal childhood intro
Prompt: seedance2 Refer to the storyboard in @image1: shots, scale, camera moves, visuals, and copy. Create a 15s healing opening about the four seasons of childhood.
Time‑lapse extension
Prompt: seedance2 15s
Sunflower scooter extension
Prompt: seedance2 10s
Cat & dog talk show
Prompt: seedance2 A roast-style dialogue in the Cat & Dog Roast Room, rich in emotion, like a stand-up performance. Cat host (licking fur, rolling eyes): Folks, who gets this? The one next to me--every day, besides wagging his tail and tearing up the sofa, only uses that I'm-so-good-pet-me look to trick humans out of snacks. When he wrecks the house he's fiercer than anyone, yet he calls himself Wangzai; I say Wang-wreck fits better, hahaha. Dog host (tilting head, wagging tail): You can talk? You sleep 18 hours a day, wake up to rub humans' legs for canned food, shed so much their black clothes are full of your fur. They just swept the floor, you turn around and roll on the sofa again--yet you still act like a cold noble?
Yuju opera voice scene
Prompt: seedance2 The accompaniment for the pre-trial section of the Yuju opera Zha Mei An begins. On the left, Bao Zheng in black points at Chen Shimei in red on the right and sings through clenched teeth: Blade to scabbard; with solid evidence, how dare you not confess? Chen Shimei's eyes dart around looking for a way out, extremely embarrassed. Then offscreen, a Yuju dan role calls: Wait! Bao Zheng and Chen Shimei look to the right.
One‑take rooftop chase
Prompt: seedance2 @image1 @image2 @image3 @image4 @image5. A one-take tracking shot follows the runner from the street up the stairs, through the corridor, onto the rooftop, ending with an overhead view of the city.
One‑take spy tracking
Prompt: seedance2 Spy thriller style. Use @image1 as the first frame. The camera front-tracks a female agent in a red coat walking forward; a wide shot follows as passers-by block her. At a corner, reference the building in @image2. Hold a fixed camera as she exits frame and disappears around the corner. A masked girl hides at the corner, glaring; her mask look references @image3. The camera pans forward to the agent; she enters a mansion and disappears; the mansion references @image4. No cuts, one take.
Singer replacement edit
Prompt: seedance2 Replace the female singer in @video1 with the male singer from @image1. Actions fully imitate the original video; no cuts; the band performs.
Shark reveal edit
Prompt: seedance2 Change the woman's hair in @video1 to long red hair. The great white shark from @image1 slowly rises with half its head behind her.
Music beat styling
Prompt: seedance2 The girl in the poster keeps changing outfits; clothing references @image1 and @image2; she carries the bag from @image3. The video rhythm references @video.
Beat‑synced landscapes
Prompt: seedance2 Scenery images @image1 @image2 @image3 @image4 @image5 @image6. Reference the rhythm in @video; sync transitions to the beat with matching visual style and music.
Mirror breakdown emotion
Prompt: seedance2 The woman in @image1 walks to the mirror and looks at herself; her pose references @image2. After thinking for a moment, she suddenly breaks down screaming and grabs the mirror. The breakdown emotion and expressions fully reference @video1.
Kitchen contrast ad
Prompt: seedance2 This is a range hood ad. Use @image1 as the first frame; a woman cooks elegantly with no smoke. The camera quickly pans right to @image2 where a man, sweating and flushed, cooks amid heavy smoke. The camera pans left and pushes in to a range hood on the table in @image1; the hood references @image4 and is powerfully sucking smoke.
Model overview
seedance2 multi-modal control with reference-anything
seedance2 combines flexible inputs and reference control to keep style, motion, and camera consistent across shots.
Multi-modal inputs
Text, image, video, and audio are all supported in seedance2.
Reference anything
Use seedance2 to reference style, subject, motion, camera, or sound.
Consistency boost
seedance2 keeps characters, outfits, products, and typography stable.
Motion & camera replication
seedance2 follows complex action rhythms and camera language.
Video extension & editing
seedance2 can extend clips or edit existing footage with references.
Audio generation
seedance2 can generate dialogue, ambience, and SFX with video.
seedance2 multi-modal workflow for text, image, video, and audio
Follow the seedance2 workflow to build controllable text-to-video, image-to-video, and video-to-video clips with reference-anything control and audio sync.
- 01/ 03
1) Add prompt + multi-modal assets
Combine text prompts with up to 9 images, 3 videos (≤15s total), and 3 audio files (≤15s total); set 4–15s duration.
- 02/ 03
2) Assign @ references
Use @ to tag first/last frame, style, subject, motion, camera, or audio roles. Choose First/Last-frame for simple I2V or Omni-Reference for full mixes.
- 03/ 03
3) Generate, extend, and edit
Preview assets, then generate, extend, or edit shots to keep character, motion, and camera consistency across sequences.
seedance2 prompt best practices
Use these seedance2 tips to keep references, motion, and audio aligned.
Input limits & duration
seedance2 supports multimodal inputs with clear caps.
- Images ≤ 9; videos ≤ 3 (≤15s total); audio ≤ 3 (≤15s total).
- Total files ≤ 12; duration 4–15s.
Choose the right mode
seedance2 supports First/Last-frame and Omni-Reference.
- If only a single image + prompt, use First/Last-frame.
- For multi-asset mixing, use Omni-Reference and @ tags.
Declare references explicitly
In seedance2 prompts, state what each asset controls.
- Example: @image1 as first frame; @video1 for camera/motion; @audio1 for rhythm.
- Check @ names to avoid mix-ups.
Structure shots
Write subject → action → camera → style.
- Add shot size + movement (push/pull/follow).
- Use Shot 1/2/3 for multi-shot timing.
Audio cues
Include dialogue tone, ambience, and SFX.
- Short lines help lip-sync.
- Name the sound source (wind, crowd, footsteps).
Extension & editing
Tell seedance2 how long to extend and what to replace.
- Example: “extend @video1 by 5s”.
- For edits, specify target (character/prop/scene).
Output settings
seedance2 can output higher resolutions when available.
- Up to 2K output on supported configurations.
- Choose platform-friendly aspect ratios and formats.
FAQ
FAQ
Get started
Start creating with Seedance 2
Use the generator above: write a prompt (and optionally upload inputs), tune params, then generate.