Seedance 2.5.

Four things changed meaningfully. First, native clip length jumped from around 10 seconds to 30 seconds in a single generation pass (and 180 seconds in beta), so you get a complete scene rather than a fragment. Second, the reference cap went from 12 to 50 multimodal inputs, which is enough to load a full production brief. Third, region editing is new: you can fix one zone of the clip without re-rolling everything. Fourth, the resolution ceiling moved from 1080p to 4K, and prompt adherence improved by 20% over 2.0. R2V motion control was in 2.0 but is more precise in 2.5.

Yes. Seedance 2.5 generates up to 30 seconds of video in a single pass, holding character, lighting, and narrative momentum across the full duration without stitching separate clips together. In beta mode the limit extends to 180 seconds. The result is a clip with consistent continuity throughout, not three 10-second fragments assembled in post.

Region editing lets you highlight a specific area of the generated clip — a background, a face, a foreground object — and re-render only that part. The rest of the clip stays exactly as it was generated. Think of it as a professional retake of one element without reshooting the whole scene. You describe the change for the selected zone, and Seedance 2.5 applies it while leaving everything outside the region untouched.

On MITO, Seedance 2.5 sits inside a canvas where all 50 references are already loaded and travel with every generation. On a direct API or standalone tool, you reconstruct that context each time. MITO also gives you region editing, storyboards, and scene continuity next to the generation, so a 30-second clip can become a sequence and a sequence can become a finished piece without switching tools.

Up to 50 multimodal references per generation: images (character sheets, location boards, style guides), video clips (for camera or character motion via R2V), and audio tracks. You point to each reference by @tag in your prompt, so the model knows which elements to draw from which input. This is up from 12 in Seedance 2.0.

Up to 4K, which is four times the resolution ceiling of Seedance 2.0 (1080p max). The higher resolution makes it viable for large-format screens, print-to-video workflows, and platforms that reward sharpness. Aspect ratios range from 9:16 (vertical) to 21:9 (ultrawide).

R2V (Reference-to-Video) lets you upload a video clip that demonstrates the camera move or character path you want. Seedance 2.5 reads the motion logic from that reference and applies it to your scene. If your reference shows a slow 180-degree orbit around a subject, the model transfers that exact arc — its speed, timing, and framing — to your new scene. It gives you precise camera direction with greater fidelity than text description alone.

Seedance 2.5, Kling 3.0, and Runway each have a different strength. Seedance 2.5 leads on deliverable completeness: 30-second native generation, 50 references, region editing, and 4K output make it the strongest choice when you need a finished clip in one pass. Kling 3.0 leads on natural motion and physics for shorter clips where movement has to feel real. Runway excels at precise creative control and professional output quality for production work. All three are available on MITO, so you choose by what the clip needs.

Yes. Videos generated with Seedance 2.5 through MITO can be used for commercial purposes including advertising, social media campaigns, short films, and client deliverables. All generated content is yours to use. Check MITO's terms of service for full details on commercial usage rights and licensing.

Credits per generation are TBD and will be confirmed before launch. Start with free credits to run your first Seedance 2.5 generation on MITO. No credit card needed. Additional credits can be purchased through your account or included in a MITO subscription plan.