MAGI-2-preview by Sand AI: new open-source image-to-video model on HuggingFace
On HuggingFace, the model sand-ai/MAGI-2-preview has been published — a preview version of the image-to-video generator from Sand AI. The model tackles the image-to-video task, and the code is available as open source. As of the publication date of August 3, 2026, it has 34 likes and no downloads yet.
What it means
MAGI-2-preview does not directly overlap with the tools in our AI for 3D reference. All three models we track — OpenLRM, Hunyuan3D 2.1, and TRELLIS — handle the image-to-3d task, meaning they create 3D models from an image. MAGI-2-preview generates video, not 3D geometry, so there is no direct comparison with them.
Still, the release of an open image-to-video model from Sand AI expands the landscape of generative tools that take images as input. The developer does not specify VRAM requirements, export formats, or other technical details — the model is marked as preview, and there is no detailed documentation yet.How the method works. The diagram is drawn based on this news note.
New h3-vbvr model on HuggingFace tackles image-to-videoOn August 18, 2026, the Patarapoom/h3-vbvr model was published on HuggingFace with the image-to-video task. At the time of publication, it had 2,434 d…
Token Radius Attention Speeds Up Video Generation in Diffusion Transformers Without Fine-TuningResearchers from Peking University have introduced Token Radius Attention (TRA), a method that reduces the computational load in Video Diffusion Trans…
GVCCTurbo proposes bitrate planning for generative compression of video and images withoutarXiv Researchers introduced GVCCTurbo, a planner that separates costly updates of the generative prior from transmitting corrections through a codebo…
ROAD cuts the cost of 3D generation training by transferring knowledge from discriminative modelsThe authors of ROAD: Reciprocal-Objective Alignment of Discriminative Semantics for 3D Shape Generation propose a method that uses discriminative 3D m…