Seedance 2.5
Direct a full 30-second scene in 4K with up to 50 references on Seedance 2.5. Describe the shot in plain language, and the agent writes the prompt that gets your scene made.
What's new in Seedance 2.5
Longer scenes in a single pass
Seedance 2.5 doubles the length of a generation. A whole 30-second scene can play out in one shot, instead of being assembled from shorter pieces.
Up to 50 references in one brief
A single generation now accepts up to 50 references, three times what Seedance 2.0 allowed. Every recurring character, product, and object in the scene can be pinned to a reference instead of a description.
3D inputs for composition
Seedance 2.5 accepts a 3D white-model input, so the layout of a shot can be staged in rough 3D first, then rendered out. It gives you control over composition before you commit to the final look.
Native 4K outputs
Generations run up to 4K, so your finished scene is delivery-ready without needing a separate upscale pass.
Precision editing
One wrong element no longer costs you the whole generation. Region-level editing re-draws a specific part of the frame while everything else stays exactly as it was.
How Seedance 2.5 compares to Seedance 2.0
Capability | Seedance 2.5 | Seedance 2.0 |
|---|---|---|
Clip length | 30 seconds, single native pass | Up to 15 seconds |
Reference inputs | Up to 50 | Up to 15 (9 images, 3 video, 3 audio) |
Prompt adherence | ~20% better | Baseline |
Precision editing | Yes | Not available |
3D white-model input | Yes | Not available |
Resolution | Up to 4K (3840×2160) | Up to 4K |
How Seedance 2.5 will work with invideo agents
Seedance 2.5 will run inside the same agentic workflow every model does on invideo: you direct in plain language, and the agent does the building. Here is what that will look like.
You describe the shot, the agent structures the prompt
Seedance 2.5 rewards structured prompting, scene setup, then action, then a closing frame. Instead of doing that yourself, you describe the shot in plain language and the agent shapes your direction into the structure the model reads best.
Your characters and locations travel with every generation
The agent attaches your approved characters and locations to every shot. With room for up to 50 references, Seedance 2.5 holds every recurring subject, no matter what you generate.
You always stay in control
How much the agent runs on its own is your setting: let it generate, ask before video, or show every step first. That choice is yours to make and yours to change.
What Seedance 2.5 is built to do best
Long single-take scenes
A 30-second generation holds a full beat: an establishing move, a dialogue exchange, a continuous piece of action. No cutting it into shorter clips and stitching them back together.
Scenes with multiple subjects
With up to 50 references in one brief, a crowded shot can pin every character and product to its own image. The cast and the props stay themselves across the whole take.
Product and brand films
You can now make a full 30-second ad in a single pass. With native audio and 4K outputs, the entire ad lands already synced and delivery-ready. And when one element is off, like a label, a logo, or a hand, region editing fixes just that without regenerating the shot around it.
Previs and composition tests
Generating complex action sequences with 3D white-model inputs on Seedance 2.5 can help you evaluate a scene before committing to an actual shoot.
Helping creatives stay creative
Multiplayer mode
Collaborate in real time with live cursors to show what everyone's working on.
Storyboarding
Turn any script or idea into a shot-by-shot plan, then tweak as needed before generating.
Script writing
Write your script inside invideo, and ask an AI co-writer for help if you'd like.
Timeline editor
Picture Premiere Pro with full AI.
Build your own agents
Create custom agents to fill specific roles like cinematographer, music designer, and more.
From solo creatives to creative enterprises
World-class investors stand behind invideo.
Backed by the firms behind Stripe, Spotify, Flipkart, and ByteDance.
Pricing
Access to 200+ image, video, audio, music models including Seedance 2.0, Veo 3.1, Kling 3.0, Nano banana pro & Elevenlabs music.
Access to top stock providers like iStock, Storyblocks & more.
Model & agent prices are subject to change.
On-demand credit top-ups available.
Seedance 2.5 FAQs
What is Seedance 2.5?
Seedance 2.5 is ByteDance's next-generation AI video model. It generates a 30-second video in one pass, in 4K, from up to 50 references. It adds precision region editing and 3D white-model input for composition. On invideo, you direct it in plain language and the agent runs it for you.
What's the difference between Seedance 2.5 and Seedance 2.0?
Clip length doubles, from 15 seconds to 30. References go from 15 to 50. Prompt adherence improves about 20%. Region editing lets you fix part of a frame without a full re-render. And 3D white-model input lets you block composition first. On invideo, the agent picks whichever model fits the shot.
Does Seedance 2.5 really generate a 30-second clip in a single take?
Yes. Seedance 2.5 generates a full 30 seconds in one native pass, no stitching shorter clips together. On invideo, the agent runs that generation for you, and chains takes into a longer sequence when you need one.
Will characters stay consistent across a full 30-second clip?
Yes. On invideo, the agent locks your characters and locations in your project's memory and carries them into the generation. With up to 50 references, every recurring face, outfit, and product stays itself, start to finish. No mid-clip identity resets.
Can I fix part of a clip without regenerating the whole thing?
Yes. Seedance 2.5 supports precision region editing. Point the agent at the part that's off, a hand, a label, one object, and it redraws just that region. The rest of the take stays exactly as it was, so one flaw doesn't cost you a whole 30-second render.
What generation modes will Seedance 2.5 support?
Text-to-video and image-to-video. Region editing lets you change specific parts of a clip. References can be images, video, or audio. On invideo, the agent drives all of it from your direction.
How does Seedance 2.5 handle audio?
Audio is generated in the same pass as the video. Action and sound come back already in sync. There's no separate audio step. On invideo, the agent handles it all from your direction.
Do I need to write prompts for Seedance 2.5 on invideo?
No. You describe the shot in plain language. The agent writes the prompt Seedance 2.5 reads best, attaches your references, and runs it. You direct; the agent handles the machinery.
When will Seedance 2.5 be available on invideo?
Seedance 2.5 is coming soon to invideo. It joins the agent's model roster at launch. From there you direct it like any other model, in plain language, with your characters and locations attached.
Does Seedance 2.5 replace shot-by-shot control?
No. You still direct and approve every shot. The agent executes, generates takes, runs precision edits, and chains shots, while you make the calls. Longer takes are an option, not a loss of control.