ByteDance releases Seedance 2.5: XCMG, Xiaopeng, etc. will be the first to access it

Kuai Technology reported on July 31 that today, ByteDance officially released Seedance 2.5, a new generation video generation model. The new model has been fully upgraded in terms of video generation duration, multi-modal understanding, local editing accuracy and multi-language support. For enterprise users, Volcano Engine will launch the Seedance 2.5 API service in the near future to further promote the implementation of video generation capabilities in industries such as industry, automobiles, and robotics. Officials stated that many companies including XCMG, Xpeng Motors, Zhiyuan Robot, Lingchu Intelligent, Fenfen Zhifei, Qiongche Intelligent, and Xspark AI (Unbounded Intelligent Aviation) have reached cooperation intentions with Volcano Engine and will be the first to access Seedance 2.5. At present, some companies have started exploring actual scenarios based on the Seedance series models. Among them, XCMG Group applies it to industrial operation training and process SOP video generation, which can reduce the time and cost required for traditional real-life shooting and 3D animation production. Xpeng Motors has connected Seedance capabilities to its internal one-stop AI design platform to assist in visual creation and solution presentation during the product design stage. Compared with Seedance 2.0 released earlier this year, Seedance 2.5 mainly upgrades four core capabilities. First of all, the duration of a single video generation is increased to 30 seconds, and high-fidelity timing extension is supported, which can maintain the consistency of characters, scenes and lens styles in longer video content. Secondly, the new model supports up to 50 multi-modal reference materials, and adds white model, green screen and other reference capabilities, making complex projects and industrial video production more controllable. In terms of editing capabilities, Seedance 2.5 adds a new video partial editing function, which can accurately adjust local elements such as background, products, and characters while maintaining the overall picture rhythm and content structure. In addition, the new model natively supports more than 10 languages, further enhances the ability to understand and execute complex instructions, and can meet the needs of mass production of multi-language and multi-version content. With the launch of API services, Seedance 2.5 is expected to further reduce