Question 1

What is Seedance 2.5?

Accepted Answer

Seedance 2.5 is ByteDance's video generation model, released on July 31, 2026. It makes up to 30 seconds of video and matching sound in a single pass, accepts up to 50 reference files per request, generates speech natively in 11 languages, and supports editing down to the second.

Question 2

What changed from Seedance 2.0?

Accepted Answer

Four things, per ByteDance's own comparison. Single pass length doubled from 15 to 30 seconds. Timestamps now work: 2.0 only responded to shot numbers, 2.5 responds to whole second timestamps. Aspect ratio went from six fixed options to any ratio between 0.4 and 2.5. And multi view images now work as subject references, which 2.0 advised against.

Question 3

How long can a Seedance 2.5 video be?

Accepted Answer

Up to 30 seconds in one generation. Past that you use extension, where each pass continues the previous clip. ByteDance recommends MOV in and out for extension, which is what keeps colour, brightness and audio matched across the join.

Question 4

How many reference images can Seedance 2.5 use?

Accepted Answer

Up to 30 images at up to 4K each, plus 10 video clips and 10 audio clips, capped at around 30 seconds combined for each of video and audio. That is 50 files in one request. Results are most stable with 1 to 8 distinct subjects across the images, and 1 to 5 subjects for video or audio references.

Question 5

How do I reference an image in a Seedance 2.5 prompt?

Accepted Answer

Number assets by upload order and bind each one explicitly in the text, for example "Image 1 depicts the protagonist and uses the voice timbre from Audio 1". Say what each asset is a reference for rather than just naming it. ByteDance specifically advises against labelling assets inside the image itself, since writing a name on a photo tends to produce duplicate or confused characters.

Question 6

Can I set the aspect ratio and duration?

Accepted Answer

For reference, storyboard and keyframe tasks, yes. Editing, extension and first frame tasks lock the output to the file you fed in: editing locks both ratio and length, while extension and first frame lock the ratio and leave the length to you. You cannot reframe or restretch a clip through an edit pass.

Question 7

How do I stop it adding subtitles or music?

Accepted Answer

Ask for it directly. The guide keeps prompts positive everywhere else, but supports negative control for subtitles and audio: "No subtitles", "No BGM; generate only environmental sounds and action sounds", or "No audio". For brand work this is the difference between a usable cut and one you cannot ship.

Question 8

Can I use what I generate commercially?

Accepted Answer

What you make in Rendley is yours to use, including for commercial projects, marketing and social media. Check the terms of whichever platform runs the generation too, and treat generated footage like any other asset when it makes a product claim or shows a real person.

Seedance 2.5 AI video generator

Some examples, inputs included

Text only, in three blocks

Eight product photos become a vlog

A nine panel storyboard, shot for shot

Six keyframes, one vertical promo

A grey blockout becomes a finished short

Recast and relocate an existing clip

Add five seconds to the end of a shot

Translate the dialogue, keep the mouth in sync

Learn how to prompt Seedance 2.5 efficiently.

Explore other models

Seedance 2.0

Veo 3.1

Sora 2

Kling v2.6

Seedance 2.5 questions, answered

More AI video models