MiniMax-H3-slim released on HuggingFace for text-to-video generation
On HuggingFace, on August 24, 2026, the model fkyyy/MiniMax-H3-slim was published with the text-to-video task. At the time of publication, it had 508 downloads and zero likes. Details are on the model page.
What it means
The model belongs to the text-to-video task, whereas our AI for 3D and video section on 3D generation describes tools with the image-to-3d task: OpenLRM, Hunyuan3D 2.1, and TRELLIS. None of them solve the text-to-video task, so MiniMax-H3-slim cannot be directly compared with them.
According to the HuggingFace page, the developer does not specify VRAM requirements, export formats, or license. The code is released as open source, but the launch parameters and model limitations are not disclosed in the description.
For those looking for 3D generation tools, MiniMax-H3-slim is not a substitute: it generates video, not 3D assets. If you need text-to-video specifically, it is worth waiting for more detailed documentation from the model author.How the method works. The diagram is drawn based on this news note.
Luce introduces 3D asset generation with PBR materials and relighting from a single imageResearchers published on arXiv the paper Luce: Relightable Gaussians for 3D Asset Generation, presenting a method for generating 3D models from a sing…
Robust-WAM adds semantic robustness to World-Action Models built on video generatorsResearchers from arXiv presented Robust-WAM, a post-training method for World-Action Models built on video generative models. The main problem the wor…
SQuad accelerates video generation in Wan 2.2, reducing attention by 67xThe authors of SQuad introduced a sub-quadratic attention distillation method for Video Diffusion Transformers. Instead of training a model from scrat…
LTX-Video, CogVideoX and Pyramid Flow fall short of simulators on eight criteriaThe authors of a systematic review on arXiv analyzed 200 papers on generative world models from 2018 to June 2026 and compared them with traditional s…