Seedance 2.5
Experience ByteDance's next-generation multimodal video model. Generate seamless native 30-second 4K (10-bit color) video in a single diffusion pass. Create more precise, cinematic results with up to 50 multimodal reference inputs and joint audio-video generation.
🌀 Native 30s · 4K 10-bit · 50 reference inputs
Seedance 2.5 in Action
Explore real examples that show the power of multimodal AI video generation. Each demo shows how images, videos and prompts combine to create stunning results.
Creative Templates & Complex Effects
Precisely replicate fish-eye lens effects, flash overlays and outfit changes. Reference a model's appearance across multiple clothing styles with dynamic camera cuts.
“Reference the model's facial features from the first image. The model wears the outfits from reference images 2-6 and approaches the camera with playful, cool, cute, surprised and stylish poses. Each outfit change triggers a camera cut, using the fish-eye lens effect and flash overlay from the reference video.”
Motion & Camera Replication
Combine a character's motion from one video with the camera movement from another. Create an epic battle scene with dust effects and cinematic tension under a starry night sky.
“Reference the character motion from Video 1 and the orbiting camera movement from Video 2. Generate a battle scene between Character 1 and Character 2 under a starry night sky, with white dust rising during the fight. The battle is spectacular and intense.”
Native 30s Video Extension
Extend smoothly up to 30 seconds in a single diffusion pass while keeping everything consistent. Create a quirky ad in which a donkey riding a motorcycle moves through a series of cinematic shots.
“Reference @image1 and @image2 of a donkey riding a motorcycle and generate a 30-second video. Scene 1: a side shot of the donkey bursting through a fence and startling the chickens. Scene 2: the donkey doing motorcycle stunts on sand, a close-up of the tires, then an aerial shot. Scene 3: the donkey jumps with mountains in the background, and the text 'Inspire Creativity, Enrich Life' is revealed with a masking effect.”
Cinematic Audio & Visuals
Create MV-quality content with precise cinematography keywords. Build film-like compositions with golden-hour lighting and a film-grain look.
“Generate a 30-second MV. Keywords: stable composition, smooth push-pull, low-angle hero shot, premium documentary feel. Ultra-wide establishing shot, slightly upward-tilted low angle, a cliffside dirt road with a vintage travel car in the lower third, distant sea and horizon for depth, sunset side-back light with volumetric rays through dust particles, cinematic framing, authentic film grain, clothes moving gently in the wind.”
Video Editing & Character Replacement
Replace a character while keeping every action and movement from the original video. Swap a female vocalist for a male singer while the band keeps playing.
“Replace the female lead singer in Video 1 with the male singer from Image 1. His movements must match the original video exactly. No camera cuts. The band keeps playing.”
One-Take Long Shot
Create a seamless one-take sequence from multiple reference images. Follow a runner from the street, up the stairs and through a hallway to the rooftop in a single continuous shot.
“@image1 @image2 @image3 @image4 @image5, one-take tracking shot following a runner from the street up the stairs, through a hallway and onto the rooftop, ending on a view of the city.”
3D White Model Previsualization
Block out scene layout, camera framing and motion with 3D white models before the final render. With minimal input, the AI completes an emotional story.
“Use the audio from Video 1 to create an emotional video inspired by Images 1-5. Pre-define the camera path with 3D white models, and reference @video1 for the background music.”