New Minimax_h3_hybrid Model on HuggingFace Solves Image-to-Video

The model abakanai/Minimax_h3_hybrid, designed for image-to-video generation, has been published on HuggingFace. The model appeared on August 9, 2026, and at the time of publication has 2111 downloads and 6 likes. The developer does not specify the architecture, hardware requirements, or output format. The model weights are distributed under the MiniMax H3 Community License: commercial use is permitted with annual revenue up to 20 million USD, the license does not apply in the EU, the UK, Korea, and the US, using the outputs of MiniMax H3 to improve other AI models is prohibited, and the license and the NOTICE file must be included when redistributing.

What it means

This news concerns the image-to-video task, not 3D generation. The CPU3D reference currently has no tools for video generation — all the models presented there solve the image-to-3d task. The closest tools by input data — OpenLRM, Hunyuan3D 2.1, and TRELLIS — take an image as input, but output 3D models rather than video. None of them overlaps with Minimax_h3_hybrid in output format. The Minimax_h3_hybrid model expands the landscape of generative tools, but so far has no counterparts in our reference. If the developer discloses the details — VRAM requirements, license, output format — it will be possible to assess its place among content creation tools more precisely.