Video model · Live
Seedance 2.0 AI video generator
Generate from a written brief, from one or two frames, or from a mixed set of image, video and audio references — with the model locked to Seedance 2.0 for the whole conversation.
Seedance 2.0 workspace
The media mode and model are locked; the Agent reads your instruction and the verified source media to choose text-to-video, image-to-video, or reference-to-video.
Seedance 2.0 specifications
Every value below is read from the runtime model catalog that routes and prices the request, not maintained a second time in marketing copy.
- Output
- Video
- Supported modes
- Text-to-Video, Image-to-Video, Reference-to-Video
- Duration
- 4–15s
- Resolutions
- 480p, 720p, 1080p
- Aspect ratios
- adaptive, 21:9, 16:9, 4:3, 1:1, 3:4, 9:16
- Generated audio
- Supported
- Images per reference operation
- 9
What Seedance 2.0 does here
The page exposes only the operations and settings Ideart AI can route and validate for this model today.
Text-to-video
Generate a clip from a written brief alone, with nothing attached.
Image-to-video from one or two frames
Use one image as the opening frame, or two as an explicit opening and ending frame.
Reference-to-video
Guide identity, style, motion or sound with up to 9 images, 3 videos and 3 audio files, 15 references in total.
Up to 1080p
480p, 720p and 1080p are offered and 720p is the default — the highest tier either Seedance model reaches here.
Generated audio
The runtime requests generated audio alongside the picture, so a result arrives with sound to review.
Four to fifteen seconds
Clip length is chosen before generating, anywhere from four up to fifteen seconds.
How a Seedance 2.0 clip gets made
Treat the first render as a direction check, then change one variable at a time.
Describe one shot
State the subject, the action, the setting, the camera behaviour, the pacing and the sound you expect.
Say what your media is for
Name each attachment's role in plain language — an opening or ending frame, or a reference for identity, style, motion or sound — and the Agent picks the matching route.
Trim references before you attach them
A reference request is charged on the longer of your input video and the clip it produces, so a trimmed source is a cheaper one.
Set the delivery format
Choose duration, 480p, 720p or 1080p, and the aspect ratio the finished clip has to fit.
Prompts to send to Seedance 2.0
Open a clip to read the prompt beside it, then send that prompt straight into the locked workspace above. These clips are creative references from Ideart AI's media library, published so the prompt structure can be studied and reused; they are not presented as verified Seedance 2.0 outputs.
Writing Seedance 2.0 prompts
This model returns picture and sound together, so both are worth directing.
Say what changes
Describe what is different at the last second compared with the first, instead of naming a mood and stopping.
Name one camera move
Pick a single primary move — push-in, orbit, tracking, crane, or locked-off — and let the subject carry the rest.
Direct the sound
Ambience, footsteps, weather or a specific effect are worth one clause; without direction the audio is a guess.
Where Seedance 2.0 fits
Work that has to finish at 1080p, arrive with sound, or stay recognisably the same across several shots.
Finish an environment shot at 1080p
Take a wide, depth-driven shot to 1080p, the highest tier this model offers, rather than upscaling a smaller render afterwards.
- 480p, 720p and 1080p tiers
- 21:9 and 16:9 framing
- Four to fifteen second cuts
Keep a character recognisable across shots
Attach the images that define a character and describe the new action, instead of rewriting the description from scratch each time.
- Up to 9 reference images
- Mixed image, video and audio references
- 15 reference files in total
Get picture and audio in one pass
Review timing, texture and ambience together on the first take instead of judging a silent clip and scoring it later.
- Generated audio with the video
- Ambience and effects directed in the prompt
- One review pass instead of two
Seedance 2.0 next to Seedance 2.5 and MiniMax H3
The table is generated from the same runtime settings the generation service uses, so it stays aligned as operations and routes change.
| Specification | Seedance 2.0 | Seedance 2.5 | MiniMax H3 |
|---|---|---|---|
| Output | Video | Video | Video |
| Supported modes | Text-to-Video, Image-to-Video, Reference-to-Video | Text-to-Video, Image-to-Video, Reference-to-Video, Video Edit, Video Extend | Text-to-Video, Image-to-Video |
| Duration | 4–15s | 4–30s | 5–15s |
| Resolutions | 480p, 720p, 1080p | 480p, 720p | 768P, 2K |
| Aspect ratios | adaptive, 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 | auto, 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 | adaptive, 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 |
| Generated audio | Supported | Supported | Not supported |
| Images per reference operation | 9 | 30 | 1 |
Read before you use an output
Generative video is probabilistic. Watch and listen to the whole clip rather than judging one frame.
- Faces, hands, small text, repeated objects and reflections can drift between frames.
- Reference media guides a result; it does not guarantee pixel-level identity, timing or composition.
- A reference request accepts video up to 30 seconds per file and 60 seconds in total, and is charged on the longer of the input and the output.
- Input length is measured server-side from the file header. A container that cannot be read is charged at a fixed full-clip length rather than nothing.
- Seedance 2.0 has no video edit or extend route; changing or continuing an existing clip is a Seedance 2.5 job.
- Generated audio still needs review for timing, intelligibility, brand safety and the context it will be used in.
Seedance 2.0 questions
Which modes can the Agent use with Seedance 2.0?
Three: text-to-video, image-to-video from one or two frames, and reference-to-video from a mixed set of files. Editing an existing clip or extending one routes to Seedance 2.5 instead.
How much reference material can one request carry?
Up to 9 images, 3 videos and 3 audio files, with 15 files in total. Each reference video can run to 30 seconds and the videos together to 60 seconds.
How long and how large can one clip be?
Four to fifteen seconds, at 480p, 720p or 1080p, in any of the seven listed aspect ratios. 720p is the default. What each combination costs is on the pricing page, not restated here.
Are the clips on this page Seedance 2.0 outputs?
No. They are labelled creative references from the project's media library, published so the prompts beside them can be read and reused.
Why is this page not indexed yet?
The page stays noindex until its first-party examples and publication review are complete. The route works; it simply does not advertise unfinished evidence to search engines.
Start a Seedance 2.0 clip
Write one shot, say what any attached media is for, choose the delivery format, and refine the useful take in your own private conversation.