We develop AI tools, custom nodes, and AI models for cinematic production and video processing.
Improve video resolution and fine details with MiniMax H3 technology featuring temporal chunking, spatial tiling, and refinement for efficient upscaling.
Optimize attention performance for NVIDIA GB10 Grace Blackwell systems using native sparse attention, configurable quality and speed profiles, and automatic compatibility fallback.
Enhance visual quality with anti-aliasing, motion-aware stabilization, and edge-preserving detail enhancement to achieve cleaner, smoother, and more consistent video output.
Download LoRA models and workflows for AI video and image production from Hugging Face.
| LoRA Models |
Custom LoRA models for MiniMax H3, LTX-2.3/2.5, and FLUX.2 Klein 9B. Covers stylized 3D animation, cinematic scenes, product advertising, character creation, historical environments, and realistic image generation.
Browse LoRA models |
| Workflows |
ComfyUI workflows for MiniMax H3 and LTX-2.5. Includes text-to-video, image-to-video, first/last-frame animation, reference-guided generation, audio support, ControlNet motion guidance, and video upscaling and refinement.
MiniMax H3 · LTX-2.5 |


