Seedance AI Video Generator
Use Seedance to build motion-rich videos from prompts, images, or reference clips when you need scene coherence, camera direction, and cinematic framing.
Seedance 2.0 for cinematic AI video
Seedance 2.0 is the primary high-quality video generation model with strong motion coherence, multi-shot storytelling, and cinematic framing. Paired with Seedance 2.0 Fast for rapid iteration, you can explore, draft, and finalize — all within a single workflow.
Core strengths
- Turns written scenes into coherent motion with camera direction, visual rhythm, and multi-shot continuity.
- Seedance 2.0 delivers up to 1080p quality; Seedance 2.0 Fast offers 720p drafts at lower cost for rapid iteration.
- Supports text, image reference (up to 9), video reference (up to 3), and web search — alone or combined.
Best for
- Short cinematic clips, product motion studies, social video concepts, and mood explorations.
- Production teams that need to compare high-quality output against faster draft generation.
- Video projects where camera movement, lighting continuity, and scene flow are critical.
Generation settings
Prompting tips
How to use Seedance
- Step 1Describe one clear scene with subject, action, camera movement, and mood.
- Step 2Upload reference images or videos when you want stronger visual or motion guidance.
- Step 3Choose the Seedance variant, ratio, resolution, and duration for your output.
- Step 4Generate a draft, review motion quality, and refine the prompt or references.
Prompting guide
Text to Video — Ad Copy Generation
Turn product descriptions into cinematic motion
Seedance 2.0 transforms written ad copy into short, compelling video clips. Structure your prompt like a commercial script — identify the hero product, describe the scene and key action, set the mood with lighting and color direction — and the model generates polished video ready for social media, ads, or product showcases.
Based on the reference image of the high-end smartwatch, generate a 5-second commercial ad video. The camera starts at medium shot, slowly pushes forward, then subtly orbits around the product, revealing the watch face, metal bezel, crown, and strap details. Keep cool blue and silver-gray tones with a minimal dark studio backdrop and refined surface reflections. Let the product rotate slowly with stable, smooth camera movement that highlights the precision industrial design, tech feel, and luxury advertising quality. The overall style is like a luxury tech brand commercial — clean visuals, restrained pacing.
A high-end smartwatch reference shot — dark studio, cool blue and silver-gray tones, refined product photography style.
5-second commercial video generated from the reference image — slow push-in and subtle orbit around the watch with cinematic lighting.
Image Reference — Multi-View Subject
Guide video output with reference images
Upload reference images showing a subject from multiple angles — front, side, back — and Seedance 2.0 will generate video that respects the provided views. This is ideal for character animation, product demos, and any scenario where visual fidelity to the source subject matters.
Based on the woman in reference image 1, image 2, and image 3, generate a video of her sitting in a cafe eating cake.
Front, side, and back views of a woman — character reference sheet style for multi-view video generation.
Video generated from the 3 reference views — woman eating cake in a cafe, natural consistent appearance.
Video Reference — Camera Movement Guide
Control motion and framing with video input
Upload a reference video with your desired camera movement, and Seedance 2.0 will adopt that motion language while applying it to your scene. Combine a reference image for visual context and a reference video for motion — the model matches the camera choreography to your new content.
Using the camera movement from reference video 1, create a concept video of a tech campus. Use the high-rise building from reference image 1 as the visual centerpiece, with the same first-person dive-down perspective, capturing the tech-forward atmosphere of the campus shown in image 1.
Concept video combining the building from the reference image with the dive-down camera movement from the reference video.
Related tools
Community showcase
Frequently asked questions
What kind of AI videos is Seedance good for?+
Seedance is well suited to videos with visible motion, camera movement, and strong visual atmosphere, such as product motion shots, short-form social ideas, ad concepts, mood clips, and cinematic draft sequences.
What input types does Seedance support?+
The current page supports text prompts and can also work with reference images or reference videos. Reference images help guide the visual direction, while reference videos are more useful when you want to influence motion or pacing.
What is the difference between Seedance 2.0 and Seedance Fast?+
In general, the standard version is better for higher-quality output, while the Fast version is better for rapid tests and direction finding. A common workflow is to test the idea quickly first and then switch to the higher-quality version for final output.
What video settings does Seedance support?+
The current page supports common video ratios, multiple resolution options, and a limited duration range. Different Seedance variants may support different maximum resolutions or speed profiles, so the current generator settings should be treated as the source of truth.
Does Seedance support web search?+
Yes, but it should be understood carefully. On the current site, web search is more suitable for text-only video requests. If you are already using reference images or reference videos, the exact available combinations should follow the page controls.
How should I write prompts for Seedance?+
Write prompts like shot directions instead of just keywords. In addition to the subject and scene, describe movement, camera motion, pacing, atmosphere, lighting, and the visual style you want. A single clear scene usually produces more coherent video than stacking many disconnected ideas.
Can I generate an image first and then turn it into a video?+
Yes. That is a practical workflow. You can first generate a reference image on an image model page and then use it on the Seedance page as an input, which makes it easier to keep subject direction and visual tone more consistent.