Download
1 variant available
This checkpoint includes a config file, download and place it along side the checkpoint.
Type
Workflows
Stats
260 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 9
Reviews
No reviews yet
Published
Sep 16, 2026
Base Model
Other
Hash
AutoV2
6425CA0DBF
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
Show more

6380 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 9
3.4K0 1 2 3 4 5 6 7 8 9.0 1 2 3 4 5 6 7 8 9K
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

