About this model
Seedance 2.5 Flash R2V is the reference-conditioned entry point into ByteDance's Flash video tier, released in November 2026. Instead of working from a prompt alone, it ingests up to ten reference images along with reference videos and audio, then synthesizes clips from four to thirty seconds at resolutions up to 720p. It can also edit or extend existing footage, making it useful for iterating on a shot rather than regenerating from scratch each time.
Within ByteDance's lineup, it sits alongside the text-driven Seedance 2.5 Flash, the image-conditioned variant, and Seedance 2.5 Flash Edit, sharing the same Flash backbone but specializing in multi-reference input. It parallels Seedance 2.5 R2V in the standard tier, trading some headroom for quicker generation. The broader family also spans the Seedream image and inpainting models.
Best suited for work where visual consistency matters more than open-ended invention: holding a character, product, or set design steady across shots, matching an established style, or extending and revising existing clips with controlled guidance.
This About section is AI-generated from public sources (Claude Opus 5), with no human editing. It may contain inaccuracies — verify critical details against the sources listed above.
Data sources: Venice API · HuggingFace · Wikipedia — enrichment updated 2d ago