Seedance 2.0
ByteDance · AI videoSeedance 2.0 is ByteDance's next-generation video model: it generates video with synchronized audio from text, images, video clips, and audio in one pass, using a unified multimodal architecture. Provider details and examples: https://replicate.com/bytedance/seedance-2.0
Compared with earlier Seedance releases, 2.0 emphasizes multimodal references (multiple images, videos, and audio in one job), stronger motion and physics, optional video editing and extension workflows, and flexible duration and aspect handling on the provider side. It is aimed at creators who want native audio–video sync, dialogue in prompts (quoted speech for lip sync), and more control over consistency across shots.
On VidMachine, Seedance 2.0 is integrated via Replicate for project scene generation. We use 480p output at 9:16 to align with short-form delivery and pricing. Credits are 14 per second of requested scene duration, same family as our other per-second video models.
Key features and benefits
Multimodal inputs (provider)
On Replicate, Seedance 2.0 can combine multiple reference images, video clips, and audio; prompts can reference them as [Image1], [Video1], [Audio1], etc. VidMachine's automated pipeline uses text-to-video and image-to-video (start frame, optional last frame) like our other models—advanced multi-asset references are available on the provider for custom workflows.
Native synchronized audio
Audio and video are generated together, so dialogue, effects, and music stay aligned with picture. You can describe speech in double quotes in prompts for lip-synced delivery. The provider also supports turning audio off for silent video if needed.
Editing and extension (provider)
The model supports describing edits to an existing clip or continuing a reference video with consistent characters and style. These flows are documented on ByteDance and Replicate; VidMachine focuses on standard scene generation from screenplay and frames.
Resolution on VidMachine
We call the model at 480p for vertical shorts (496×864 at 9:16 per ByteDance's published resolution table). This keeps generation cost predictable while matching mobile-first formats.
Technical specifications
ProviderReplicate — bytedance/seedance-2.0 (https://replicate.com/bytedance/seedance-2.0)
VidMachine resolution480p, 9:16
Scene duration2–12 seconds per clip (aligned with Seedance 1.5-style limits in our integration)
Credits14 credits per second of requested duration
Use cases and applications
Seedance 2.0 suits short-form social, ads, and narrative clips where you want native audio, dialogue, and motion in one generation. It is a strong option when you outgrow simpler text-only video and want ByteDance's latest multimodal model in your priority stack.
Use it alongside Seedance 1.5 Pro or other models in video model priority: if the primary model fails or is slow, fallbacks keep automation moving.
Why this model
Choose Seedance 2.0 when you want ByteDance's newest unified audio–video model with multimodal capabilities on the provider, at a clear 14 credits per second and 480p vertical output on VidMachine.
If you need 720p Seedance today, Seedance 1.5 Pro remains available in the same picker with its own per-second rate.
How VidMachine uses it
Select Seedance 2.0 in your project's video model priority (up to three models). We generate scenes through Replicate with prompt, optional start image from your pipeline, optional last frame when the workflow supplies it, 480p, 9:16, and audio enabled.
Credits deduct per second of target scene length for the seedance-2-0 rate in project settings. Check Pricing for the latest numbers.
What you should know
Is Seedance 2.0 available on VidMachine?
Yes. It is available as seedance-2-0 in video model priority, powered by Replicate's bytedance/seedance-2.0 model.
Why 480p?
VidMachine requests 480p for Seedance 2.0 to match your requirement for that tier and to keep credit usage predictable for high-volume shorts.
Where are official specs and limits?
See https://replicate.com/bytedance/seedance-2.0 and ByteDance's Seedance 2.0 documentation for authoritative limits, resolutions, and prompt patterns.