/
#54
Tencent Hy
Tencent's open-source Hunyuan model family covering text-to-image (HunyuanImage, diffusion-based, up to 80B MoE), text/image-to-video (HunyuanVideo, 13B+ DiT with Causal 3D VAE), and text/image-to-3D (Hunyuan3D, flow-matching DiT + texture synthesis). Models are Diffusers-compatible; inference supported via xDiT for parallelism.
Categories
Subcategories
VIDEO GENERATIONVIDEO UNDERSTANDINGWORLD MODELS
Links