HiT trains end-to-end from pixels, hits 1.48 FID with 2.4x less compute
CatAstro_Piyush · x · 2026-11-19
HiT is an image generator trained end-to-end from pixels without a separately trained encoder. It learns its own hierarchy of abstractions, dynamically grouping patches across multiple levels—semantic structure emerges without masks or region labels. In the demo video, backgrounds collapse into a few groups, the main subject gets many, and words light up the regions they refer to.
Performance: 1.48 FID on ImageNet 256, 2.4x less training compute than JiT-H/16 at matched quality, and it outperforms models with 6x more parameters on average across text-to-image benchmarks. Website, paper, and code are released.
More from Multimodal
- Blogger predicts next Fable video models could closely clone GTA 6-level graphics with enough tokens — ChrisGPT · 2026-11-19
- Crumpled-paper websites imagined in Nano Banana: the hard part is now the idea — danshipper · 2026-11-19
- DeepMind launches Genie 4 interactive world model with real-time physics, reasoning and audio — demishassabis · 2026-11-19
- Free ComfyUI workflow runs P-H3-Image at half MiniMax H3's size on 6GB VRAM — TheLocalLab · 2026-11-19
- Open-source Claude Code plugin turns any song into a finished music video automatically — AIandDesign · 2026-11-19
- Procedural 3D will be the backbone of video generation, not pure video models — dreamwieber · 2026-11-19