--- title: GenRF emoji: 🖼️ colorFrom: indigo colorTo: blue sdk: gradio sdk_version: 5.49.1 app_file: app.py python_version: "3.12.12" pinned: false short_description: Registration-free image fusion with generative priors --- [Paper](https://github.com/Linfeng-Tang/GenRF) · [Code](https://github.com/Linfeng-Tang/GenRF) · [Project page](https://linfeng-tang.github.io/GenRF/) # GenRF Image Fusion Choose an infrared–visible, multi-exposure, or multi-focus example to inspect the two inputs and its archived result. GPU instances also support live inference on two uploaded images of matching dimensions. Input pairs with a long edge over 1280 px are downsampled at their original aspect ratio. The app shows the selected prompt and uses 30 denoising steps, guidance 3.5, and seed 2026. Live inference uses the GenRF LoRA and the gated FLUX.3 Image Klein Base model. Configure the Space with a read-only HF_TOKEN secret with access to the base model and CUDA hardware. The examples remain available on a CPU Space.