NextFin News — JD.com on Wednesday open-sourced its self-developed real-time streaming video editing model JoyAI-Video-Edit under the Apache 2.0 license.
The model lets users modify characters and scenes while a video plays, turning traditional post-production workflows into interactive, concurrent editing. Natural-language instructions support adding or removing objects, replacing people, transferring clothing styles and reconstructing entire scenes without requiring a finished source clip first. The system extends JD’s broader JoyAI multimodal suite, which already includes image-editing, long-form audio-video generation and real-time vision-language interaction models released earlier in 2026.
Code and weights are available on the company’s public GitHub organization, enabling developers to integrate the streaming capabilities into creative tools, e-commerce content pipelines and live-production applications. The release continues JD’s pattern of opening foundation models that link generation with immediate, instruction-driven refinement across visual media.
Explore more exclusive insights at nextfin.ai.
