Skip to content
Synced · WeChat

Video AI Moves to 5 Minutes: Fully Open Source, One-Pass Generation, No Blind-Box Sampling

视频AI卷向5分钟:全量开源,一次生成,正式告别「盲盒抽卡」

JD open-sourced JoyAI-Echo, a long audio-video generation framework that supports up to 5 minutes of cross-shot audiovisual consistency, local repainting, 8-step DMD distillation, and output up to 1472×2560 resolution.

Why it matters: JoyAI-Echo clears HKR-H/K/R with a concrete open-source long-video claim: 5-minute output, cross-shot audio-video consistency, and 8-step DMD distillation. Single-source coverage and no independent evals keep it in the 78–84 band.

Read the original ↗Export Markdown