Generate
Edit
Customize
"People running, risograph poster"
Generate
Edit
Customize
Generate a complete 30-second ad with opening, product moment, and closing frame in a single generation. Characters, lighting, and scene continuity stay locked from start to finish.
Open Krea VideoSupply a reference for every character, product, or object in your scene. With up to 50 multimodal inputs, Seedance 2.5 holds each subject stable across the full clip.
Open Krea VideoFix a product label, adjust a face, or replace a background by editing only that region. The rest of the clip stays intact, preserving lighting, composition, and motion.
Open Krea VideoPrompt Examples
Seedance 2.5 holds characters, lighting, and native synchronized audio steady across a full 30-second generation — no stitching, no separate sound pass. Every example below is paired with the exact prompt that generated it.
Prompt
A young woman with windswept dark hair and an oversized brown hoodie glances backward over her shoulder. Striking sunlight cuts across her face in golden-hour warmth, strands of hair lifting in the wind. Slow push-in on a handheld 35mm frame, shallow depth of field, subtle film grain. Ambient street noise and a soft gust of wind synchronized with the hair movement.
Try this promptA young woman with windswept dark hair and an oversized brown hoodie glances backward over her shoulder. Striking sunlight cuts across her face in golden-hour warmth, strands of hair lifting in the wind. Slow push-in on a handheld 35mm frame, shallow depth of field, subtle film grain. Ambient street noise and a soft gust of wind synchronized with the hair movement.
Try this promptPrompt
A pale pink Victorian house stands in dense night fog, glowing with warm light from curtained windows. Slow dolly forward along the front path, volumetric fog catching the glow, cool moonlit rim light playing against the warm interiors. Distant wind chimes and muffled night ambience rise gently as the camera approaches the porch.
Try this promptA pale pink Victorian house stands in dense night fog, glowing with warm light from curtained windows. Slow dolly forward along the front path, volumetric fog catching the glow, cool moonlit rim light playing against the warm interiors. Distant wind chimes and muffled night ambience rise gently as the camera approaches the porch.
Try this promptPrompt
Floating modern soccer cleat with a beige textured upper and a translucent sole plate with traction studs, clean white studio background. The cleat rotates slowly in zero gravity while macro details resolve: stitching, texture grain, light refracting through the sole. Orbiting camera ends in a snap to a centered hero frame, punctuated by a clean whoosh and a click on the lock-in.
Try this promptFloating modern soccer cleat with a beige textured upper and a translucent sole plate with traction studs, clean white studio background. The cleat rotates slowly in zero gravity while macro details resolve: stitching, texture grain, light refracting through the sole. Orbiting camera ends in a snap to a centered hero frame, punctuated by a clean whoosh and a click on the lock-in.
Try this promptPrompt
A person sleeps in a bed facing a large window overlooking a vibrant neon-lit cityscape at night. The camera drifts slowly from over the bed toward the glass, city lights pulsing softly across the sheets. Rain streaks the window; distant traffic and a low synth hum give the scene a calm, cyberpunk mood.
Try this promptA person sleeps in a bed facing a large window overlooking a vibrant neon-lit cityscape at night. The camera drifts slowly from over the bed toward the glass, city lights pulsing softly across the sheets. Rain streaks the window; distant traffic and a low synth hum give the scene a calm, cyberpunk mood.
Try this promptPrompt
Extreme close-up of dual square headlights emitting blinding white light on a red and grey retro vehicle. The lights flare on with a mechanical clunk, lens flares blooming across chrome trim. Slow pull-back reveals wet asphalt reflections while an engine rumble builds beneath an electronic ignition chime.
Try this promptExtreme close-up of dual square headlights emitting blinding white light on a red and grey retro vehicle. The lights flare on with a mechanical clunk, lens flares blooming across chrome trim. Slow pull-back reveals wet asphalt reflections while an engine rumble builds beneath an electronic ignition chime.
Try this promptHow It Works
From first prompt to production-ready video in four steps.
Step 1
Go to the Krea Video tool and pick Seedance from the model selector. Seedance 2.0 is available today; Seedance 2.5 will appear in the same selector when it goes live.
Step 2
Upload reference images or clips to lock characters, products, and style. Seedance 2.5 raises the limit to 50 multimodal references, including video and audio.
Step 3
Describe the shot like a director: subject, camera movement, lighting, mood, and the sounds that should sync to the action. Longer, more specific prompts give the model more to adhere to.
Step 4
Generate, compare takes, then enhance and upscale the best one — everything stays inside Krea, ready to download for social, ads, or client work.
Seedance 2.5 is ByteDance's latest AI video generation model. It produces cinematic 30-second clips in up to 4K resolution, accepts up to 50 multimodal reference inputs, and delivers 20% better prompt adherence compared to Seedance 2.0.
Seedance 2.5 is rolling out to Krea soon. You can generate with Seedance 2.0 on Krea today — sign up and you will be notified as soon as Seedance 2.5 goes live.
Seedance 2.5 extends clip length to 30 seconds (up from 15), increases resolution to 4K, supports up to 50 reference inputs (up from 12), improves prompt adherence by 20%, and adds region-level editing so you can fix parts of a video without regenerating the full clip.
Region-level editing lets you select a specific area of a generated video and re-render only that part. You can replace objects, adjust faces, or change backgrounds while the rest of the clip stays intact.
Yes. Seedance 2.5 generates audio in the same pass as the video. Footsteps, ambient sounds, and effects are synchronized to on-screen action without a separate audio step.
Seedance 2.5 generates clips up to 30 seconds in standard mode and up to 4K resolution. An extended beta mode supports clips up to 180 seconds.
Up to 50 multimodal references including images, video clips, and audio files. This allows precise control over character appearance, scene composition, and motion direction in complex multi-subject scenes.
Yes. On Krea paid plans, generated videos include commercial licensing for campaigns, products, client work, and other production projects.
Generate cinematic 30-second AI videos in up to 4K resolution with 20% better prompt adherence. Combine up to 50 multimodal references, use region-level editing to refine without regenerating, and produce clips with native audio sync.
Try Seedance on KreaExplore other ByteDance models available on Krea
Seedream 5 Lite
Medium quality model with reasoning and web search.
Seedream 4.5
Medium quality model for photorealism.
Seedance 2.0
Flagship video model with native audio, references, and frame animation.
Seedance 2.0 Fast
Faster Seedance 2 with native audio, references, and frame animation.
Seedance 2.0 Mini
Cost-efficient Seedance 2 model for scalable video production.
Discover more AI models for images and videos on Krea
