本人插件github:https://github.com/yusu-02/ComfyUI-Yusu-MiniMaxH3-Unified
dlss插件:https://github.com/RH-RunningHub/ComfyUI-RH-DLSS5
Youtube:https://www.youtube.com/@2002yusu
bilibili: https://b23.tv/FP8uE0o
--------------runninghub新人福利码------------
国内用户邀请码:https://www.runninghub.cn/?inviteCode=cn-v1084
海外用户邀请码:https://www.runninghub.ai/?inviteCode=rh-v1525
--------------本期视频线上体验区------------
鱼酥的AI站:https://vibex.runninghub.cn/p/app-b3d967d5d3d44cc8a9ddd789af586cbc/?inviteCode=rh-v1525
Minimax工作站:https://yusu.apps.vibex.cn/?inviteCode=rh-1525
国内用户: minimaxH3三采黑科技工作流:https://www.runninghub.cn/post/2104527536003833858/?inviteCode=cn-v1084
国外用户: minimaxH3三采黑科技工作流:https://www.runninghub.ai/zh-cn/post/2105310704311304193/?inviteCode=rh-v1525
--------------本期模板+模型+插件领取区------------
本期网盘链接:https://pan.quark.cn/s/8b91fb71fb3d
超清工作流网盘链接:https://pan.quark.cn/s/ff5b49e4d4a8
超清软件:https://pan.quark.cn/s/362b780e0d9d
往期网盘链接:https://pan.quark.cn/s/f79e125ffc22
这是一套面向 MiniMax H3 的综合视频生成与 4K 超分工作流,整合了 黑科技 SelfLift 工作流 和 常规单采工作流 两种生成方式,可以根据素材和硬件情况自由选择。工作流支持文生视频、图生视频,并加入加速 LoRA 和画质 LoRA,在提升生成效率的同时,改善 H3 常见的油腻感、高光偏亮和画面质感问题。
生成完成后,工作流提供多种视频后处理和超分路线,包括 DLSS、VOSR、FlashVSR 以及快速 VAE/latent 超分方案,可根据速度、细节、时序稳定性和最终分辨率选择不同方案。工作流本身划分了独立的 DLSS、VOSR、FlashVSR 和快速超分区域。
其中 VOSR 和 FlashVSR 都提供了独立的视频源切换,可以在 黑科技工作流视频、普通单采视频、DLSS 处理后视频 三种来源之间直接选择,不需要重新连接节点,方便快速测试和比较不同的 4K 输出组合。
整体上,这套工作流更像一个完整的 MiniMax H3 视频工具箱:既可以快速生成基础视频,也可以根据实际素材,自由组合不同生成方式和超分方案,兼顾生成速度、画面质感、细节表现和 4K 输出需求。
This is an all-in-one MiniMax H3 video generation and 4K enhancement workflow that combines two generation pipelines: a SelfLift high-resolution workflow and a standard single-pass H3 workflow. Users can choose between them depending on the source material, hardware, and desired output. The workflow supports both text-to-video and image-to-video generation, while combining acceleration and quality-oriented LoRAs to improve efficiency and reduce common H3 issues such as overly oily textures, harsh highlights, and excessive brightness.
After generation, multiple video enhancement and upscaling options are available, including DLSS, VOSR, FlashVSR, and a fast VAE/latent-based upscaling route. These options are designed for different priorities such as processing speed, detail recovery, temporal consistency, and final output resolution.
Both the VOSR and FlashVSR sections include a built-in three-way source selector, allowing users to switch directly between the SelfLift output, standard single-pass output, and DLSS-enhanced output without reconnecting nodes manually. This makes it easy to test and compare different 4K processing pipelines.
Overall, the workflow works as a complete MiniMax H3 video production toolkit, covering generation, quality enhancement, flexible upscaling, and multiple 4K output routes in a single workflow.
Description
这是一套面向 MiniMax H3 的综合视频生成与 4K 超分工作流,整合了 黑科技 SelfLift 工作流 和 常规单采工作流 两种生成方式,可以根据素材和硬件情况自由选择。工作流支持文生视频、图生视频,并加入加速 LoRA 和画质 LoRA,在提升生成效率的同时,改善 H3 常见的油腻感、高光偏亮和画面质感问题。
生成完成后,工作流提供多种视频后处理和超分路线,包括 DLSS、VOSR、FlashVSR 以及快速 VAE/latent 超分方案,可根据速度、细节、时序稳定性和最终分辨率选择不同方案。工作流本身划分了独立的 DLSS、VOSR、FlashVSR 和快速超分区域。
其中 VOSR 和 FlashVSR 都提供了独立的视频源切换,可以在 黑科技工作流视频、普通单采视频、DLSS 处理后视频 三种来源之间直接选择,不需要重新连接节点,方便快速测试和比较不同的 4K 输出组合。
整体上,这套工作流更像一个完整的 MiniMax H3 视频工具箱:既可以快速生成基础视频,也可以根据实际素材,自由组合不同生成方式和超分方案,兼顾生成速度、画面质感、细节表现和 4K 输出需求。
This is an all-in-one MiniMax H3 video generation and 4K enhancement workflow that combines two generation pipelines: a SelfLift high-resolution workflow and a standard single-pass H3 workflow. Users can choose between them depending on the source material, hardware, and desired output. The workflow supports both text-to-video and image-to-video generation, while combining acceleration and quality-oriented LoRAs to improve efficiency and reduce common H3 issues such as overly oily textures, harsh highlights, and excessive brightness.
After generation, multiple video enhancement and upscaling options are available, including DLSS, VOSR, FlashVSR, and a fast VAE/latent-based upscaling route. These options are designed for different priorities such as processing speed, detail recovery, temporal consistency, and final output resolution.
Both the VOSR and FlashVSR sections include a built-in three-way source selector, allowing users to switch directly between the SelfLift output, standard single-pass output, and DLSS-enhanced output without reconnecting nodes manually. This makes it easy to test and compare different 4K processing pipelines.
Overall, the workflow works as a complete MiniMax H3 video production toolkit, covering generation, quality enhancement, flexible upscaling, and multiple 4K output routes in a single workflow.
FAQ
Comments (4)
I would downvote if we had the function for it.
what are we supposed to do with a full chinese workflow with custom nodes that are not included or link provided inside the workflow?
This is sloppy, confusing and honestly not something worth sharing.
It's not like dude has access to the world's best translations tools to have ever exist!
It doesn't work. half the thing needs nodes that you cannot find.