Open-Source AI Workflow for Digital-Human Talking-Head Videos
阿星 published an open-source workflow for producing digital-human talking-head videos with WorkBuddy and flashV4, covering topic selection, scripting, filming, effects packaging, cover creation and publishing. His reference setup combines interchangeable tools including LibTV MCP, Seedance 2.0, MiniMax, Wan 3.0 and Astra, with koubo-video-script and koubo-video-package skills handling scripting and packaging. The author says he tested the workflow himself but notes deformation or blur in the character, unstable cloned speech and occasional lip-sync misalignment.
The report provides a concrete, configurable production pipeline with practical steps and openly shared skills, while documenting the quality issues users may need to review manually.
哈喽,大家好
我是阿星👋🏻上一篇讲的是WorkBuddy做科普视频Skill:中文解说配音字幕全包
这一篇讲我真正每天在做的事:
基于WorkBuddy+flashV4低成本完成口播视频动效包装,比opus基于cc少点细腻但可以出品小白可上手。下面这条就是👇
除了人物有点变形了模糊了声音需要再调整下,基本没毛病。
我把整套流程拆成了 5 步:选题、脚本、拍摄、包装、封面和发布。
#WorkBuddy#flashV4#koubo-ai-skills#koubo-video-script#koubo-video-package#LibTV MCP#Seedance 2.0#MiniMax#Wan 3.0#OmniHuman 2.0#DeepFilterNet#Whisper