<div data-youtube-video="https://youtu.be/bXFhG_OlEUU"><iframe src="https://www.youtube.com/embed/bXFhG_OlEUU" width="853" height="480" allowfullscreen></iframe></div> 
 
AI-KSK ComfyUI workflow: MiniMax H3 TaoMate 3 Step Acceleration LoRA. 
 
This workflow is the TaoMate 3-step acceleration final version from the 1080P near-1-frame-per-second benchmark. It runs a two-pass pipeline on the MiniMax H3 hybrid FL2VA/Ref2VA bf16 model: Pass1 uses the TaoMate-H3 step3000 LoRA with a 3-step schedule for fast draft generation, then Pass2 refines with SelfLift acceleration and the full wushu LoRA stack. The graph is a multi-branch master layout with several bypassed branches; only the active branch produces the final 2K output. 
 
Main features: 
- TaoMate-H3 step3000 ComfyUI-BF16 acceleration LoRA on the model path 
- 3-step simple schedule per active generation branch (euler) 
- MiniMax H3 hybrid FL2VA/Ref2VA bf16 model with Qwen3-VL 32B INT8 ConvRot encoder 
- Wushu LoRA stack: wushu_e3 FL2VA (0.75), H3_Combat_V2 (0.35), camera motion (0.35), spatial physics (0.3), Motion Repair (0.2), weapon combat (0.1) 
- Pass2 SelfLift refinement into the 2K output 
- Audio VAE decode, MP4 video saved at 24 fps 
 
Suggested workflow: 
Use this workflow as the fast final-version route for digital human talking videos and wushu action shots. Run the 3-step pass first to check identity and motion direction, then let the refinement pass finish the frame. Keep the reference face stable and one clear action sentence per prompt; some duplicate branches in the graph are bypassed by design, so do not enable them manually. 
 
RunningHub workflow (try online, no install needed): 
https://www.runninghub.ai/zh-cn/post/2100147178424578050?inviteCode=rh-v1111 
 
AI-KSK RunningHub workflow hub: 
https://www.runninghub.ai/user-center/1892566146468511746?inviteCode=rh-v1111 
 
Bilibili tutorial: 
https://www.bilibili.com/video/BV1MYec6uEoN/ 
 
RunningHub registration and AI tools: 
https://www.runninghub.ai/?inviteCode=rh-v1111 
https://www.runninghub.ai/ai-generator/2046514150500524034?inviteCode=rh-v1111 
https://www.runninghub.ai/vip-rights/3?inviteCode=rh-v1111 
https://rhtv.runninghub.ai/projects?inviteCode=rh-v1111 
 
Stable global access: 
https://lelians.com/#/register?code=mFIJ5tZn 
 
TokenRhythm AI model credits: 
https://tokenrhythm.studio/i/rf_tr_kM6wekDquDDZQD4PUNmgiuHA 
 
Youyun Cloud GPU: 
https://passport.compshare.cn/register?referral_code=A7fTQUYY9lCDFJwA6MMvoq 
https://www.notion.so/2b72459a3ac6808f8282f2c687ed8f52 
 
RunPod Cloud GPU: 
https://runpod.io/?ref=sxk5n5m7 
 
Support AI-KSK: 
https://ko-fi.com/aiksk

MiniMax H3 TaoMate 3 Step Acceleration LoRA

<div data-youtube-video="https://youtu.be/bXFhG_OlEUU"></div> 
 
AI-KSK ComfyUI workflow: MiniMax H3 TaoMate 3 Step Acceleration LoRA. 
 
This workflow is the TaoMate 3-step acceleration final version from the 1080P near-1-frame-per-second benchmark. It runs a two-pass pipeline on the MiniMax H3 hybrid FL2VA/Ref2VA bf16 model: Pass1 uses the TaoMate-H3 step3000 LoRA with a 3-step schedule for fast draft generation, then Pass2 refines with SelfLift acceleration and the full wushu LoRA stack. The graph is a multi-branch master layout with several bypassed branches; only the active branch produces the final 2K output. 
 
Main features: 
- TaoMate-H3 step3000 ComfyUI-BF16 acceleration LoRA on the model path 
- 3-step simple schedule per active generation branch (euler) 
- MiniMax H3 hybrid FL2VA/Ref2VA bf16 model with Qwen3-VL 32B INT8 ConvRot encoder 
- Wushu LoRA stack: wushu_e3 FL2VA (0.75), H3_Combat_V2 (0.35), camera motion (0.35), spatial physics (0.3), Motion Repair (0.2), weapon combat (0.1) 
- Pass2 SelfLift refinement into the 2K output 
- Audio VAE decode, MP4 video saved at 24 fps 
 
Suggested workflow: 
Use this workflow as the fast final-version route for digital human talking videos and wushu action shots. Run the 3-step pass first to check identity and motion direction, then let the refinement pass finish the frame. Keep the reference face stable and one clear action sentence per prompt; some duplicate branches in the graph are bypassed by design, so do not enable them manually. 
 
RunningHub workflow (try online, no install needed): 
https://www.runninghub.ai/zh-cn/post/2100147178424578050?inviteCode=rh-v1111 
 
AI-KSK RunningHub workflow hub: 
https://www.runninghub.ai/user-center/1892566146468511746?inviteCode=rh-v1111 
 
Bilibili tutorial: 
https://www.bilibili.com/video/BV1MYec6uEoN/ 
 
RunningHub registration and AI tools: 
https://www.runninghub.ai/?inviteCode=rh-v1111 
https://www.runninghub.ai/ai-generator/2046514150500524034?inviteCode=rh-v1111 
https://www.runninghub.ai/vip-rights/3?inviteCode=rh-v1111 
https://rhtv.runninghub.ai/projects?inviteCode=rh-v1111 
 
Stable global access: 
https://lelians.com/#/register?code=mFIJ5tZn 
 
TokenRhythm AI model credits: 
https://tokenrhythm.studio/i/rf_tr_kM6wekDquDDZQD4PUNmgiuHA 
 
Youyun Cloud GPU: 
https://passport.compshare.cn/register?referral_code=A7fTQUYY9lCDFJwA6MMvoq 
https://www.notion.so/2b72459a3ac6808f8282f2c687ed8f52 
 
RunPod Cloud GPU: 
https://runpod.io/?ref=sxk5n5m7 
 
Support AI-KSK: 
https://ko-fi.com/aiksk

Files with SHA256: 6425ca0dbf0f24fdfca59fa0020f57de7dffe7fd157ad3bcb3df338dc9be1fc9