Dreamina Seedance 2.5 on VM0. One-pass 30-second video
ByteDance's latest Seedance video model. Build longer, reference-controlled clips with native audio in a single generation.
Video / Text-to-video / Image-to-video / Audio
Dreamina Seedance 2.5 is ByteDance's long-form, reference-controlled video model. It generates 4 to 30 second clips at 480p or 720p, can create native audio, and accepts up to 50 reference assets in one request: 30 images, 10 videos and 10 audio clips. First-frame and last-frame controls are available too.
Compared with Dreamina Seedance 2.0, the 2.5 route doubles the maximum clip length and handles a much larger multimodal brief. The trade-off is resolution and cost: 2.0 reaches 1080p and is the better fit for shorter, cheaper iterations, while 2.5 is built for complete 30-second stories and reference-heavy production.
What is Dreamina Seedance 2.5?
August 2026 · Long-form, reference-controlled tier of ByteDance's Seedance video family.
BytePlus released Dreamina Seedance 2.5 in August 2026 as the next step beyond short-clip generation. The headline changes are native 30-second generation, up to 50 multimodal references, stronger temporal consistency and optional generated audio.
VM0 exposes the model through the dreamina-seedance-2.5 alias. The current route supports text prompts, image, video and audio references, first and last frames, every duration from 4 to 30 seconds, and 480p or 720p output.
What's notable about Dreamina Seedance 2.5
Headline architecture and capability features.
Multimodal video generation model for text-to-video and reference-guided video. A request can combine text with images, videos, audio, and boundary frames, then return one continuous clip with optional generated audio. Usage is billed from output video tokens, with a lower token rate when the input includes video references.
Specs at a glance
Dreamina Seedance 2.5 pricing
Vendor list price per generated unit.
How Dreamina Seedance 2.5 behaves in practice
Observed behaviour from production agent runs.
30-second continuity
The longest native single-pass duration in VM0's curated video lineup. It can carry scene changes, camera movement and pacing through one 30-second clip without stitching multiple generations.
Multimodal reference control
Up to 50 references let one brief hold product angles, characters, environments, motion examples and audio cues together. This is the model's clearest advantage for brand-controlled production.
Native audio
Generated audio can ship in the same pass as the video, while reference audio can guide rhythm and sound. That removes a separate audio-generation step for many social and campaign clips.
Resolution ceiling
Output tops out at 720p on the current VM0 route. Dreamina Seedance 2.0 reaches 1080p, while Veo 3.1 Fast and Kling V3 4K are the alternatives when final resolution matters more than duration and reference count.
Best agent tasks for Dreamina Seedance 2.5
The complete 30-second campaign spot
Generate the opening, product story and closing beat in one pass. Longer continuity reduces the visible seams and repeated setup work that come with stitching short clips.
The reference-heavy product video
Combine packaging shots, product angles, brand scenes, motion references and a soundtrack in one brief. The larger reference budget keeps more of the art direction inside the generation request.
The controlled first-to-last-frame transition
Set both boundary frames, add image or video references for the middle, and generate a continuous transition with optional audio across any duration from 4 to 30 seconds.
When to skip Dreamina Seedance 2.5
Skip Dreamina Seedance 2.5 when 1080p or 4K delivery is a hard requirement, when a short low-cost draft is enough, or when repeatable seed-based reruns matter. Use Dreamina Seedance 2.0 for 1080p and cheaper short clips, Veo 3.1 Fast for short cinematic output up to 4K, or Kling V3 4K for stylised native 4K video.
Dreamina Seedance 2.5 vs other models
Dreamina Seedance 2.5 vs MiniMax H3
Seedance 2.5 leads on duration and reference capacity, with clips up to 30 seconds and 50 multimodal references. MiniMax H3 leads on output resolution with 2K video and always-on native stereo audio, but tops out at 15 seconds and 12 reference files.
Dreamina Seedance 2.5 vs Dreamina Seedance 2.0
Seedance 2.5 doubles the maximum duration from 15 to 30 seconds and expands the reference budget to 30 images, 10 videos and 10 audio clips. Seedance 2.0 remains the better fit for cheaper short clips, seeded reruns and 1080p output.
Dreamina Seedance 2.5 vs Veo 3.1 Fast
Seedance 2.5 leads on clip length and multimodal reference control. Veo 3.1 Fast leads on short cinematic output and reaches 4K. Both can generate audio, so choose by duration, reference complexity and delivery resolution.
Dreamina Seedance 2.5 vs Kling V3 4K
Seedance 2.5 is the longer, reference-heavy option with native audio and clips up to 30 seconds. Kling V3 4K is the higher-resolution choice for stylised or display-ready 4K output up to 15 seconds.
Bottom line: should you use Dreamina Seedance 2.5?
Choose Dreamina Seedance 2.5 for complete 30-second stories, native audio and briefs that depend on many image, video and audio references. Choose another route when 1080p or 4K delivery, deterministic seeds or the lowest iteration cost matters more.
Frequently asked questions
What duration and resolution does Dreamina Seedance 2.5 support on VM0?
Every whole-second duration from 4 to 30 seconds, at 480p or 720p. The current VM0 route defaults to 720p.
How many references can one request include?
Up to 30 image references, 10 video references and 10 audio references, for 50 multimodal assets in total. You can also provide first and last frames.
Does Seedance 2.5 generate audio?
Yes. Generated audio is supported, and audio clips can also be supplied as references for rhythm, sound and timing.
How is Dreamina Seedance 2.5 priced on VM0?
Usage is measured in output video tokens. VM0 charges $8.00 per 1M output tokens when the input includes video and $13.375 per 1M without video input. The effective per-second cost varies with resolution and the request's input mix.
Does the VM0 route expose Seedance 2.5's region-level editing?
Not currently. VM0 exposes generation with multimodal references, first and last frames, and optional audio. Region-level editing is an upstream Seedance capability but is not part of the current built-in generation command.
Alternatives
Using Dreamina Seedance 2.5 on VM0
Using Dreamina Seedance 2.5 on VM0
VM0 agents can call Dreamina Seedance 2.5 as part of an agent run, billed against your VM0 credits. The list price above is what the upstream provider charges; VM0 passes that through with the standard credit conversion.
Available on VM0 since August 2026.