https://youtu.be/SUJkkT6XLHI 
 
This workflow combines a Larry Turbo MiniMax H3 route with a SageAttention acceleration environment. It is designed for users testing whether attention optimization can help MiniMax H3 video generation run more smoothly while keeping the graph easy to operate. 
 
The workflow keeps the familiar MiniMax H3 model, Qwen3-VL encoder, video VAE, audio VAE, scheduler, and SaveVideo structure, while the acceleration focus is the Larry Turbo plus SageAttention setup around the generation path. 
 
This package is meant for practical ComfyUI and RunningHub users who want a ready-made MiniMax H3 graph instead of rebuilding loader, encoder, VAE, sampler, and video export connections by hand. The workflow is especially useful for speed comparison, hardware planning, and repeatable prompt tests. Keep the same source image and prompt when comparing versions so the difference you see comes from the workflow route, not from changing creative variables. 
 
Main features: 
 
- MiniMax H3 acceleration workflow prepared for a SageAttention environment 
- Larry Turbo style route for faster video generation tests 
- Connected MiniMax H3 model, encoder, and VAE chain 
- 16:9 video output orientation for common tutorial and demo formats 
- Short-step sampling path for practical iteration 
- Clean graph layout for testing speed, stability, and render consistency 
- Useful when comparing attention optimization against plain turbo routes 
 
Suggested workflow: 
 
Run the same source image and prompt that you use in the Larry Turbo workflow, then compare speed, output stability, and temporal consistency. Keep the first test strict and repeatable: same resolution, same duration, same prompt, and the same audio or motion request. 
 
For better comparisons, make one small test first, review identity stability and temporal motion, then raise prompt complexity only after the base route behaves correctly. When using speaking-character prompts, keep the face visible, avoid too many fast cuts, and separate action, expression, and camera instructions clearly. 
 
RunningHub Workflow 
 
Try the workflow online right now - no installation required. 
Workflow: https://www.runninghub.ai/post/2086424118061641730?inviteCode=rh-v1111 
 
If the results meet your expectations, you can later deploy it locally for customization. 
 
Fan Benefits: Register to get 1000 points + daily login 100 points - enjoy 4090 performance and 48 GB super power! 
 
Bilibili Updates (Mainland China &amp; Asia-Pacific) 
 
If you're in the Asia-Pacific region, you can watch the video below to see the workflow demonstration and creative breakdown. 
Bilibili Video: https://www.bilibili.com/video/BV17nuU6xE3g/ 
 
Support Me on Ko-fi 
 
If you find my content helpful and want to support future creations, you can buy me a coffee. 
Every bit of support helps me keep creating. 
Ko-fi: https://ko-fi.com/aiksk 
 
Business Contact 
 
For collaboration or inquiries, please contact aiksk95 on WeChat. 
 
打开下方链接即可在线体验，无需安装。 
工作流：https://www.runninghub.ai/post/2086424118061641730?inviteCode=rh-v1111 
 
如果你觉得效果理想，也可以在本地进行自定义部署。 
 
粉丝福利：注册即可领取 1000 积分，每日登录再领 100 积分，体验 4090 和 48GB 大显存性能。 
 
B站视频（中国大陆及亚太地区） 
 
如果你在中国大陆或亚太地区，可以通过下面的视频查看工作流的实测效果与创作思路。 
B站视频：https://www.bilibili.com/video/BV17nuU6xE3g/ 
 
我会在夸克网盘持续更新模型资源： 
https://pan.quark.cn/s/07bdc81784ce

MiniMax H3 SageAttention Workflow

04_minimax-h3-sageattention-workflow.json

v1.0

<div data-youtube-video="https://youtu.be/SUJkkT6XLHI"><iframe src="https://www.youtube.com/embed/SUJkkT6XLHI" width="853" height="480" allowfullscreen></iframe></div> 
RunningHub workflow page: https://www.runninghub.ai/zh-cn/post/2086411824722956290?inviteCode=rh-v1111
<h2>MiniMax H3 Larry + Sageattention Acceleration — MiniMax H3 Larry + SageAttention Acceleration</h2>
A MiniMax H3 video graph from episode 291, part of the six-times-acceleration series that compares model tier, LoRA, attention and cache settings.
<h2>Saved configuration</h2>
91 nodes / 111 links. Four parallel branches: UNET <code>minimax_h3_fl2va_int8_convrot.safetensors</code>, CLIP <code>qwen3vl_32b_minimax_h3_int8_convrot</code>, video fp16/audio fp32 VAEs, turbo-v4 LoRA at 1, euler + beta 6 steps denoise 1; four LoadImage inputs and four outputs.
<h2>How to use</h2>
This is the Larry + SageAttention test file: four branches share the same int8 model and turbo LoRA but differ in the attention/cache patches wired around them. It exists to compare speed, not to be the cleanest single pipeline; if you want one graph, use the Larry Turbo file instead.
<h2>Limits and not-verified items</h2>
No inference was run for this listing: no speed, VRAM, quality, audio-sync or acceleration-factor claim is made. The "6x" in the episode title is the author's description and was not independently measured. Values are quoted from the saved nodes. H3 weights, LoRAs, encoder and VAEs are not bundled; check their licenses separately. The graph notes credit the official Comfy-Org/MiniMax-H3 model files; preserved as-is, no originality claim. The cover image is a promotional still, not a frame produced by this graph.
Tutorial: <a href="https://www.youtube.com/watch?v=SUJkkT6XLHI" rel="ugc">YouTube</a> · <a href="https://www.bilibili.com/video/BV17nuU6xE3g" rel="ugc">Bilibili</a>. Workflows supplied through AI-KSK; verify workflow, node, model, LoRA and cover rights separately. No blanket commercial or redistribution license is asserted.

MiniMax H3 Larry + SageAttention Acceleration

<div data-youtube-video="https://youtu.be/SUJkkT6XLHI"></div> 
RunningHub workflow page: https://www.runninghub.ai/zh-cn/post/2086411824722956290?inviteCode=rh-v1111
MiniMax H3 Larry + Sageattention Acceleration — MiniMax H3 Larry + SageAttention Acceleration
A MiniMax H3 video graph from episode 291, part of the six-times-acceleration series that compares model tier, LoRA, attention and cache settings.
Saved configuration
91 nodes / 111 links. Four parallel branches: UNET <code>minimax_h3_fl2va_int8_convrot.safetensors</code>, CLIP <code>qwen3vl_32b_minimax_h3_int8_convrot</code>, video fp16/audio fp32 VAEs, turbo-v4 LoRA at 1, euler + beta 6 steps denoise 1; four LoadImage inputs and four outputs.
How to use
This is the Larry + SageAttention test file: four branches share the same int8 model and turbo LoRA but differ in the attention/cache patches wired around them. It exists to compare speed, not to be the cleanest single pipeline; if you want one graph, use the Larry Turbo file instead.
Limits and not-verified items
No inference was run for this listing: no speed, VRAM, quality, audio-sync or acceleration-factor claim is made. The "6x" in the episode title is the author's description and was not independently measured. Values are quoted from the saved nodes. H3 weights, LoRAs, encoder and VAEs are not bundled; check their licenses separately. The graph notes credit the official Comfy-Org/MiniMax-H3 model files; preserved as-is, no originality claim. The cover image is a promotional still, not a frame produced by this graph.
Tutorial: <a href="https://www.youtube.com/watch?v=SUJkkT6XLHI" rel="ugc">YouTube</a> · <a href="https://www.bilibili.com/video/BV17nuU6xE3g" rel="ugc">Bilibili</a>. Workflows supplied through AI-KSK; verify workflow, node, model, LoRA and cover rights separately. No blanket commercial or redistribution license is asserted.

Files with SHA256: c01e661ce074c4529f9c920839dadd1b1735b8835c5b0a3f2ca7b31ceb6557f4