MiniMax-H3-Int8Convrot极速版,文图生视频,8G显存可用,自动补帧,25步采样,自定义分辨率,自适应端口,自动提示词,首尾帧,批量任务队列
单视频通过提示词直接修改


替换效果:
目标

结果

分块放大节点作者:https://github.com/bbaudio-2025/Comfyui-MMH3-UltimateUpscale

多个场景描述用 --- 分割
no music.Visual style: high-quality cinematic live-action, sharp focus, clear details, professional lighting, no blur.
[0-3s]
The beautiful and sexy super woman is kneeling on the ice with bare feet,looking at the camer
---
[3-8s]
She had an orgasm on her face and speaks in Chinese voice: <d>[Chinese] 晚安,玛卡巴卡</d>

节点作者:https://github.com/LBH-123-AI/Comfyui_Minimax_h3_latent_Upscaler


节点作者:https://github.com/kijai/ComfyUI-SolAttn_triton

采样器增强作者:https://github.com/blepping/ComfyUI-bleh

4b编码器作者:https://huggingface.co/NicoLab28/ClipProj-MiniMax-H3


默认视频清晰度:


新增新版的多运镜提示词模版

1、该镜像支持自启动,初始化后,需要等待服务启动,大概2分钟左右,可以输入命令 tail -50f /root/wan/log.txt 查看启动日志
2、随后点击 SD-WEBUI 按钮即可,上传图片,输入提示词,点击生成即可,支持批量任务



