ByteDance confirms Seedance 2.5 with longer clips and more creative control

ByteDance has confirmed the name and launch window for its next AI video generation model. Ryan Merket reports for RuntimeWire that Seedance 2.5 is set to arrive in early July, bringing three notable upgrades over its predecessor: native 30-second single-shot video output, a dramatically expanded capacity for reference materials, and improved tools for controlling and editing generated video.

What’s new in Seedance 2.5

• 30-second clips: The previous model, Seedance 2.0, topped out at 15 seconds of native output. Doubling that length is significant because maintaining visual consistency across a longer clip is one of the hardest problems in AI video.
• More reference material: Seedance 2.5 reportedly supports 50 “full-modal reference materials.” Seedance 2.0 was capped at 9 images, 3 video clips, and 3 audio clips. A higher limit means creators can supply more context in a single pass, including product shots, character references, and style frames.
• Better editing control: The update promises more precise control over generation and editing, addressing a common frustration: the need to regenerate an entire scene just to make a small change.

The information comes from posts on X (Twitter) by users @chetaslua and @alisaqqt, citing ByteDance directly. ByteDance itself has not announced pricing, API access, or which of its products will include Seedance 2.5.

Distribution will be a key question. When Seedance 2.0 rolled out to CapCut earlier this year, ByteDance introduced restrictions around real faces, unauthorized intellectual property, and mandatory watermarking. A model that accepts more references and generates longer clips raises those concerns further.