TL;DR
ByteDance announced Seedance 2.5 at its Beijing conference, creating 30 seconds of native 4K video from 50 multimodal reference inputs.
ByteDance unveiled Seedance 2.5 at the Volcano Engine FORCE conference in Beijing on Tuesday.video generation model that produces 30-second clips at native 4K resolution from a single query. The company has skipped four intermediate versions altogether, jumping directly from its predecessor to signal what it describes as a generational leap.
The enterprise beta is now live, with a public launch scheduled for early July. CEO Liang Rubo said at the conference that rising to the top of artificial intelligence is the company’s top priority, with its business-as-a-service model becoming a core operation supported by long-term investments.
The headline upgrade is reference capacity: the model accepts up to 50 multimodal inputs, including up to 12 images, audio clips, 3D white models and style references from its predecessor. These inputs give Seedance 2.5 more control over style, movement and composition than a single text query.
The model renders natively in 4K instead of downscaling to a lower resolution, which is important for professional production pipelines. It supports 10-bit color depth for smoother gradients and more room for post-production color grading. ByteDance also claims 20 percent better operational consistency, which means fewer generations before a usable result.
Audio is now co-processed in the same hidden space as visual signals, creating local synchronization between on-screen actions and their corresponding sound effects. A new 3D white-box preview feature allows creators to create low-res animations before starting full-resolution rendering. Together, the features position the model as a production tool rather than as a generator of innovation.
The announcement comes three months after ByteDance was forced out Add watermark and IP protections to Seedance 2.0 Following cease and desist letters from Disney, Warner Bros Discovery, Paramount and Netflix. A viral deep spoof of Tom Cruise fighting Brad Pitt on a rooftop has prompted an official complaint from the Motion Picture Association and a reprimand from SAG-AFTRA.
ByteDance stopped global streaming in mid-March and didn’t resume through CapCut until the end of March, with face-blocking filters, C2PA watermarks, and copyrighted character detection. No timeline has been suggested for the new model’s US launch.
The competitive landscape has changed dramatically since February. OpenAI shut down Sora in March the video tool has grossed just over two million dollars after reaching nearly a million users and reportedly costing about a million dollars a day to operate.
Google’s Veo 3.1 fills much of the vacuum, offering native 4K output, audio generation, and up to three reference images for style control. But the new ByteDance model significantly exceeds Veo’s reference input capacity, accepting 50 inputs to Veo’s three, a gap important for professional workflows.
The AI video production market has fragmented rapidly, with Chinese models moving faster than Western competitors in production tools. third party platforms such as Reallusion’s AI Studio has already built professional pipelines around its predecessorand Runway’s fourth-generation tool Artificial Analysis rounded out the top 10.
Whether the new model can reach global markets without reigniting the copyright battles that stalled its predecessor remains a major question. ByteDance’s model features distribution through CapCut’s 400 million monthly active users and vertical integration from generation to editing to sharing. She still doesn’t have a deal with Hollywood, and every feature that makes the model more capable adds to the stakes of this unresolved conflict.
