Seedance 2.5
Create 4–30 second videos from text, first and last frames, or a mix of image, video, and audio references. Generate native synchronized sound, choose 480p or 720p, and keep up to 50 reference inputs organized in one prompt.
A longer, more controllable video model
Seedance 2.5 is the version of ByteDance's Seedance line that expands the Seedance workflow with longer clips, a much larger multimodal reference budget, and synchronized sound generated with the picture in one pass.
30 Seconds in One Pass
It generates a native 30-second clip without post-stitching. That is enough to hold a full social unit, a product story, or a short scene without cutting between separate renders.
Up to 50 Reference Inputs
It takes up to 50 reference materials in a single generation, mixing images, video, and audio. That is roughly four times the input budget of the previous model, which helps keep characters and look consistent.
Native Synchronized Audio
Generate dialogue, ambience, sound effects, and music together with the video. Audio is available in text, image, and reference generation modes.
Seedance 2.5 at a glance
Current settings available through the live Seedance 2.5 API.
Clip Length
Up to 30 seconds in a single generation, with scene changes and tempo shifts handled inside the clip rather than by joining separate renders.
Resolution
Choose 480p for lower-cost drafts or 720p for standard output. Image-to-video automatically follows the first frame's ratio.
Reference Inputs
Up to 50 multimodal references per job, combining images, video, and audio. Useful for scenes that need several characters to stay on-model across shots.
Three Input Modes
Start with text, one or two frame images, or a multimodal mix of image, video, and audio references.
Audio & Web Context
Native audio is available in every mode. Text-to-video can also use web search when a prompt needs current information.
Availability
Seedance 2.5 is available in the generator above now. Every job is saved to My Assets when it starts.
What changed since Seedance 2.0
Seedance 2.0 set the baseline and currently ranks first on the Artificial Analysis video leaderboard, ahead of Google Veo 3.1 and Kling 3.0. Seedance 2.5 builds on that foundation in three places.
Longer clips
Seedance 2.0 clips ran short, usually a single beat. Seedance 2.5 holds a 30-second arc in one render, so a hook, a turn, and a payoff can live in the same clip.
More references
The earlier model accepted around a dozen references. Seedance 2.5 raises that to 50, which makes multi-character scenes and brand-consistent campaigns far easier to keep coherent.
Synchronized sound
Seedance 2.5 can create dialogue, music, and effects alongside the picture, so sound and motion are planned together.
Where Seedance 2.5 fits
Longer clips, a larger reference budget, and native audio make Seedance 2.5 useful for complete ideas instead of isolated shots.
Brand films and ads
A 30-second spot can be one render instead of a dozen stitched fragments. Feed in product shots and brand references so the look holds from the first frame to the last.
Episodic short drama
The 50-input reference system was built with multi-character scenes in mind. It keeps the same faces and wardrobe across shots, which is the hard part of serialized content.
Dialogue-led scenes
Put spoken lines in quotes and describe the sound design in the same prompt so performance, timing, and effects stay aligned.
Reference-driven composition
Use @image, @video, and @audio labels to tell the model exactly which source controls identity, camera movement, or sound.
Seedance 2.5 FAQ
Common questions about Seedance 2.5.
What is Seedance 2.5?
Seedance 2.5 is ByteDance's next AI video model, following Seedance 2.0. Its headline change is a native 30-second clip generated in a single pass, with scene cuts and tempo changes handled inside the shot rather than stitched together after the fact.
How long can a Seedance 2.5 clip be?
Up to 30 seconds in one generation. That breaks past the short 5 to 15 second clips most video models produce and is long enough to carry a complete social post or a short scene on its own.
How many reference inputs does Seedance 2.5 support?
Seedance 2.5 accepts up to 50 reference materials in a single job, combining images, video, and audio. That is roughly four times the previous model's limit and is aimed at scenes with several characters that need to stay consistent.
What can I use as a Seedance 2.5 reference?
Reference-to-video accepts images, MP4 or MOV clips, and WAV or MP3 audio. Use @image1, @video1, and @audio1 in the prompt to assign each source a role.
What resolution and audio options are available?
The current API supports 480p and 720p output. Synchronized dialogue, sound effects, and music can be enabled in every generation mode.
How is Seedance 2.5 different from Seedance 2.0?
Seedance 2.5 extends the workflow with videos up to 30 seconds, up to 50 multimodal references, and synchronized audio across text, image, and reference generation.
Is Seedance 2.5 available now?
Yes. Use the generator at the top of this page. The credit estimate updates automatically as you change duration, resolution, or add reference-video seconds.
Can I cancel my subscription or get a refund?
Yes. You can cancel your subscription at any time. For a refund, email us from the address you registered with so we can find your account. We usually reply within 3 business days.
Start creating with Seedance 2.5
Scroll back to the generator, choose a mode, and submit your first Seedance 2.5 video. Completed results remain available in My Assets.