ByteDance has announced Seedance 2.5, an AI-powered video generation tool that can generate videos up to 30 seconds long from up to 50 reference clips.



ByteDance announced its video generation AI, ' Seedance 2.5, ' at the ' FORCE Powerhouse Conference ' hosted by its cloud business, Volcano Engine. Seedance 2.5 is characterized by its ability to take up to 50 multimodal reference materials, including images, videos, and audio, as input, and to output a video of up to 30 seconds in a single generation.

Seedance 2.5 feature list points to longer AI video clips - RuntimeWire
https://runtimewire.com/article/bytedance-confirms-seedance-2-5-for-early-july-with-30-second-ai-video

ByteDance's Seedance 2.5 breaks the 30-second barrier for AI video generation
https://the-decoder.com/bytedances-seedance-2-5-breaks-the-30-second-barrier-for-ai-video-generation/

Seedance 2.5 is the successor to Seedance 2.0, a video generation AI announced in February 2026. You can learn more about what kind of video generation AI Seedance 2.0 is by reading the following article.

TikTok's developers announce video generation AI 'Seedance 2.0,' which can input up to 9 images and up to 3 reference videos - GIGAZINE



While the previous generation Seedance 2.0 featured the generation of high-quality multi-shot videos up to 15 seconds long, Seedance 2.5 allows for a single video to be generated up to 30 seconds in length. This makes it easier to create videos that include scene changes and tempo changes in a continuous sequence, making it suitable for short advertisements, product introductions, and narrative videos.

The following is a video created by stitching together footage actually output by Seedance 2.5.



Another major feature of Seedance 2.5 is its ability to handle up to 50 reference materials. While Seedance 2.0 could generate videos by referencing images, videos, and audio in addition to text, Seedance 2.5 allows for the combination of even more materials. Furthermore, by providing reference materials for people, products, backgrounds, screen atmosphere, camera work, and audio all together, it becomes easier to maintain visual and directorial consistency even with multiple characters and scenes. However, the specific input formats and upper limits for each type of material have not been revealed as of the time of writing.



ByteDance states that Seedance 2.5 will improve not only video generation but also editing usability. While the existing Seedance 2.0 already had a mechanism to generate and edit videos according to instructions, reflecting composition, movement, camera work, visual effects, and sound based on reference material, Seedance 2.5 will allow for more detailed control, including the ability to modify content while maintaining the overall look and style of the video.



Renoise, an AI video creation service, has announced that it plans to offer Seedance 2.5 through its service, and there is a possibility that it will be expanded to services other than ByteDance after its official release.



Seedance is scheduled to officially launch in early July 2026 after global testing for enterprises, but specific release dates, pricing, API delivery methods, and available regions have not yet been announced as of the time of writing.

in AI,   Video, Posted by log1i_yk