Skip to content
QbitAI · WeChat

Chinese open-source framework targets stable 5-minute AI long-video generation

5分钟AI长视频不翻车!国产开源框架杀到全球第一梯队

JD open-sourced JoyAI-Echo, a long audio-video generation framework for 5-minute consistent videos, using cross-modal memory, DMD post-training for about 7.5x faster inference, and real-time upscaling from 720P to 1K or 2K output.

Why it matters: JD's open-source JoyAI-Echo uses a cross-modal memory module to keep characters and scenes consistent across 5-minute videos, with roughly 7.5x faster inference.

Read the original ↗Export Markdown