Seedance 2.5 Flash reference to video

Quick answer: Reference to video (R2V) means giving Seedance 2.5 Flash images, video or audio to match, not just words: up to ten images, five video clips and five audio clips per request, against fifty references on the flagship. The model keeps a face, a room or a voice from them. SceneMixer builds that bundle for every Flash segment automatically from your reference sheets and character voices.
Still from “Starfall Command”, an AI short drama generated end-to-end by SceneMixer (7 characters, 1 episode)
SampleStarfall CommandDigital HD1 episodeProduced end-to-end by SceneMixer from the scriptWatch the episodes

What each kind of reference does

  • Images: who and where — a face, a costume, a room, a prop.
  • Video: how something moves — a camera path, a gesture, a dance.
  • Audio: how something sounds — a voice, a piece of music, an effect.

Text still says what happens. The references only say what it should look and sound like.

What SceneMixer sends

In SceneMixer a Seedance 2.5 Flash segment reads the same references as the rest of the Seedance 2.5 family: up to ten images (character sheets, locations, props, character states, your own uploads) and up to five voice clips, about fourteen seconds of voice in total. That sits inside what ByteDance allows Flash per request — ten images, five videos and five audio clips.

Which sheets go with a segment follows from the @ names in its shots and its place line, so a character who is not in the segment does not take up a slot. How many reference images Flash takes

Your own images as references

On a paid plan you can upload a photo or drawing in step 2 and either have it redrawn as a reference sheet in your project’s style or use it as it is. A photo of a real person’s face may be refused by the content check. Your own reference images

R2V or image to video

Image to video starts the clip from a picture: the first frame is that image. Reference to video only borrows from the pictures, so the shot can be framed any way the text asks. For a series, references are what keep the cast the same; a start frame is for one clip. Seedance 2.5 Flash image to video

Sources

  • BytePlus on X, November 12, 2026 — the announcement, the speed claim and the USD 0.152 per second price at 720p.
  • IT Home, November 12, 2026 (Chinese) — Volcano Engine’s figures: about 0.99 yuan a second at 720p, 30-second takes, 10 images, 5 videos and 5 audio clips per request, 480p and 720p, the API date.
  • PANews — the same release in English, with the list of apps it launched in.
  • Dreamina’s Seedance 2.5 Flash page — ByteDance’s own app positions Flash for previews and frequent iteration.

Specifications are as announced on November 12, 2026 and checked on November 16, 2026; the figures for SceneMixer’s tiers on this page are read from the live price list each time the page is built.

FAQ

What is reference to video in Seedance 2.5 Flash?

Giving the model images, video or audio to match, alongside the text.

How many references can Flash take?

Up to ten images, five videos and five audio clips per request.

Does SceneMixer use references automatically?

Yes. SceneMixer attaches the sheets and voices each segment needs.

Is R2V more expensive?

No. Flash is priced per second of output.

Can I add a video as a reference in SceneMixer?

Segments use images and voice clips; video references are not part of the editor.

What is the difference from image to video?

Image to video starts from a frame; reference to video borrows the look without fixing the frame.

Let the references build themselves

Set up your cast once; every Flash take carries it.

Start a project

Updated November 16, 2026