SeeDance 2.0
What is SeeDance 2.0?
SeeDance 2.0 is the second-generation video model from ByteDance, available on Unifically as Pro, Mini, Fast, and Low Restriction variants. The headline addition over the 1.x line is omni-reference mode, where a single call accepts up to nine media inputs across images, videos, and audio, referenced inline in the prompt with @Image1, @Video1, and @Audio1 placeholders. Videos run 4 to 15 seconds. Pro generates at up to 1080p or 4K; Mini and Fast at 480p or 720p. The other two modes are familiar: text-to-video with full aspect-ratio control, and first-and-last-frame interpolation that takes a start image and an optional end image.
Key features of SeeDance 2.0
Four features cover what the model actually changes versus the 1.x line.
Omni-reference up to 9 media inputs
Mix images, videos, and audio in a single call. Reference each one in the prompt with @Image1, @Video1, @Audio1 placeholders. Audio inputs are capped at 15 seconds each. Useful for combining a look, a motion source, and a cue track in one render.
4 to 15 second clips
Wider duration range than 1.5 Pro, in single-second increments. Same range across text-to-video, first-and-last-frame, and omni-reference modes.
Pro and Fast variants on the same API
Pro outputs up to 1080p and 4K. Mini outputs at 480p or 720p at lower cost. Fast outputs at 480p or 720p and prioritises throughput. All share the same parameter surface, so switching is a one-line model ID change.
First-and-last-frame interpolation
Pass optional first_frame_url and/or last_frame_url. Aspect ratio is auto-detected from the input. Useful for product reveals and bookended brand stings.
Best for
Reference-rich creative briefs
Combine a style image, a motion clip, and an audio cue in one omni-reference call. One render produces a result that previously needed three separate passes.
Long-form social cuts
15-second clips at 1080p on Pro, in vertical or square ratios, ready for paid social slots without an edit step.
Throughput batches on Fast
Fast variant for prompt sweeps and high-volume drafts. Switch to Pro for the keepers.
First-and-last-frame product reveals
Bookend the shot with the SKU shot and the packaged shot. The model interpolates motion between them.
Reproducible generation
Seed plus the same prompt and references returns the same clip. Useful for locking a draft before you re-render at 1080p.
Style transfer with motion source
Pass a still image as the look reference and a video as the motion reference. Omni-reference combines them in a single call.
Variants
SeeDance 2.0 has five callable variants on the same API. Each one has its own model ID and price.
Pro
The higher-quality variant. Outputs at 480p, 720p, 1080p, or 4K. Use it for the final pass on hero videos, paid placements, and anything that has to land at 1080p or 4K.
Mini
The compact variant. Same three modes and omni-reference controls as Pro, but outputs at 480p or 720p. Use it for draft loops and high-volume batches when 720p is enough.
Fast
The throughput variant. Outputs at 480p or 720p and runs faster per video. Use it for prompt sweeps, draft passes, and high-volume batches before locking the keepers on Pro.
Pro Low Restriction
Same modes and resolution options as Pro, with face usage allowed and reduced content moderation.
Fast Low Restriction
Same modes as Fast, with face usage allowed, reduced moderation, and 1080p output on top of 480p and 720p.
Use cases
Run a single omni-reference call that takes a style image, a reference video for motion, and an audio cue to land a 15-second cinematic video at 1080p or 4K. Storyboard a product reveal with first-and-last-frame mode by giving the model the SKU shot and the packaged shot. Run a prompt sweep on Fast at 720p, lock the keeper with a seed, then re-render on Pro at 1080p or 4K. Generate vertical 9:16 cuts and 16:9 paid-ad cuts from the same prompt by changing the aspect ratio.
API examples
Call SeeDance 2.0 from any language by POSTing to /v1/tasks. Full parameter docs live at docs.unifically.com/models/video/bytedance/seedance-2.0-pro.
curl -X POST https://api.unifically.com/v1/tasks \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{
"model": "bytedance/seedance-2.0-pro",
"input": {
"prompt": "Cinematic slow-motion shot of a dancer performing on a rain-soaked stage",
"mode": "text_to_video",
"aspect_ratio": "16:9",
"duration": 10,
"resolution": "1080p"
}
}'
Successful submission returns a task_id. Poll GET /v1/tasks/<task_id> or set a callback_url on the request to receive the finished result.
FAQs
People also ask
SeeDance 2.0 is the second-generation SeeDance video model from ByteDance. It supports text-to-video, first-and-last-frame, and an omni-reference mode that accepts up to nine image, video, and audio references in a single call. Videos run 4 to 15 seconds, at 480p or 720p on Mini and Fast and up to 1080p or 4K on Pro.
Five. Pro is the higher-quality variant with output up to 1080p and 4K. Mini is the compact variant at 480p or 720p. Fast trades resolution for throughput, also at 480p or 720p. Pro Low Restriction and Fast Low Restriction add face usage and reduced moderation on top of Pro and Fast.
A single call can take up to nine media references (images, videos, and audio, with audio capped at 15 seconds each). Reference them in the prompt with @Image1, @Video1, @Audio1 placeholders. Useful for combining a look, a motion source, and an audio cue in one render.
4 to 15 seconds, in single-second increments. Same range across all three modes (text-to-video, first-and-last-frame, omni-reference).
It accepts audio references and can be guided by them in omni-reference mode. For native synchronized audio in the output you typically supply an audio reference; otherwise the clip is video only.
1.5 Pro is single-shot video, no references, up to 12 seconds. 2.0 adds the omni-reference mode (up to nine media inputs), pushes duration to 15 seconds, and offers a Fast variant for throughput.
Yes. Pass an integer seed alongside the prompt and references. The same prompt, references, and seed return the same render. Useful for locking in a clip you want to re-render at higher resolution on the Pro variant.
