这是一篇归档文章,最后收录于 2026-10-28 01:00(北京时间)。下方指标保留各自的采集时间,不代表当前热度。

← 返回内容列表
𝕏X采集材料

Henge:用ComfyUI制作角色替换视频

10-27 20:38 · 北京时间

图像与视频

有实践材料可继续查看

代码仓库演示线索

原文与观测指标

1 个来源
𝕏X

Henge:用ComfyUI制作角色替换视频

NRDX
查看原文
点赞61采集时的累计值
每小时增长+13.5赞 / 小时 · 观测区间平均值
观测增量+27两次有效记录之间的变化

发布时间:2026-10-27 20:38 · 北京时间

指标采集:10-28 00:00

对比区间:10-27 22:00 至 10-28 00:00 · 2.00 小时 · 净增量 +27

摘要

Henge基于ComfyUI和MiniMax H3,可用引导视频与参考图制作角色替换、动作迁移和多人换脸等视频效果。内置助手能辅助检查素材、标记角色和准备提示词,也支持本地或RunPod运行。

摘要由 AI 生成,请以原文为准。
观测历史 · 最近 2 条 展开 / 收起

缺失不计为零;增长只比较同一指标来源的有效记录,按实际间隔折算每小时。上游缓存可能延迟数值更新。

采集时间 · 北京时间指标数值来源状态
10-28 00:00:42点赞61MCP 工具连接与集成已取得
10-27 22:00:40点赞34MCP 工具连接与集成已取得
采集原文 展开 / 收起

Meet Henge — an open-source video studio named after Henge no Jutsu, the transformation technique from Naruto. Keep the performance. Transform t...

Meet Henge — an open-source video studio named after Henge no Jutsu, the transformation technique from Naruto. Keep the performance. Transform the character. 🥷 Built on ComfyUI + MiniMax H3, Henge turns a guide video and reference images into Character Swap, Motion Transfer, Head Swap, Multiple Character Swap and Sharpness/upscaling results. The existing features use LoRAs I previously released. The new Motion Transfer introduces new LoRAs I trained, including TARA, my temporal-alignment technique for following the guide’s motion. The built-in AI assistant helps inspect footage, mark and assign multiple characters, prepare prompts and launch generations faster. You can also use your own AI client through MCP. Run locally or on RunPod Pods and Serverless. Comfy Cloud integration is being validated and requires our Henge custom nodes, now published in the Comfy Registry, plus the required models and dependencies. Watch the demo below Open source. MIT-licensed app. https://github.com/alisson-anjos/Henge #OpenSourceHiggsfield