[MiniMax H3] Direct output of 2.0MP with 12GB VRAM! LoRA High-resolution workflow available
Wanting to run high resolution directly on MiniMax H3 demands too much VRAM;
Using a low-VRAM upscaling method easily results in blurring after attaching LoRA.
Verified on hardware:
- RTX 4070 Ti 12GB can directly run 2.0MP
- 10-second H3 video
- Local 12GB: About 15 minutes
- RunningHub: About 7 minutes
- Supports normal attachment of H3 LoRA
- Play high resolution without needing 24GB / 48GB graphics cards
- Core goal: Low VRAM, 2.0MP, LoRA available
For someone like me who has been constantly working on high-dynamics LoRA such as Combat, Weapon, GUNFU, and Motion Repair, this means much more than simply "being able to output 2MP".
Now, 12GB cards can finally truly enter the H3 high-resolution practical range.
And the speed is even better than I expected:
4070 Ti 12GB: About 15 minutes for 10 seconds.
RunningHub: About 7 minutes for the same task.
This solution was mainly completed by group member Xu Nuo.
H3 2.0MP is no longer exclusive to large VRAM.
12GB can handle it too.
7 minutes to direct 1080P! 12GB available_BUNNY_SelfLift_High Dynamic+LoRANo Blur_Hybrid Workflow
[MiniMax H3] Direct output of 2.0MP with 12GB VRAM! LoRA High-resolution workflow available
Wanting to run high resolution directly on MiniMax H3 demands too much VRAM;
Using a low-VRAM upscaling method easily results in blurring after attaching LoRA.
Verified on hardware:
- RTX 4070 Ti 12GB can directly run 2.0MP
- 10-second H3 video
- Local 12GB: About 15 minutes
- RunningHub: About 7 minutes
- Supports normal attachment of H3 LoRA
- Play high resolution without needing 24GB / 48GB graphics cards
- Core goal: Low VRAM, 2.0MP, LoRA available
For someone like me who has been constantly working on high-dynamics LoRA such as Combat, Weapon, GUNFU, and Motion Repair, this means much more than simply "being able to output 2MP".
Now, 12GB cards can finally truly enter the H3 high-resolution practical range.
And the speed is even better than I expected:
4070 Ti 12GB: About 15 minutes for 10 seconds.
RunningHub: About 7 minutes for the same task.
This solution was mainly completed by group member Xu Nuo.
H3 2.0MP is no longer exclusive to large VRAM.
12GB can handle it too.
[MiniMax H3] Direct output of 2.0MP with 12GB VRAM! LoRA High-resolution workflow available
Wanting to run high resolution directly on MiniMax H3 demands too much VRAM;
Using a low-VRAM upscaling method easily results in blurring after attaching LoRA.
Verified on hardware:
- RTX 4070 Ti 12GB can directly run 2.0MP
- 10-second H3 video
- Local 12GB: About 15 minutes
- RunningHub: About 7 minutes
- Supports normal attachment of H3 LoRA
- Play high resolution without needing 24GB / 48GB graphics cards
- Core goal: Low VRAM, 2.0MP, LoRA available
For someone like me who has been constantly working on high-dynamics LoRA such as Combat, Weapon, GUNFU, and Motion Repair, this means much more than simply "being able to output 2MP".
Now, 12GB cards can finally truly enter the H3 high-resolution practical range.
And the speed is even better than I expected:
4070 Ti 12GB: About 15 minutes for 10 seconds.
RunningHub: About 7 minutes for the same task.
This solution was mainly completed by group member Xu Nuo.
H3 2.0MP is no longer exclusive to large VRAM.
12GB can handle it too.
