Model link: https://modelscope.cn/models/DiffSynth-Studio/FLUX.1-Kontext-dev-lora-SuperOutpainting Editing instruction: Outpaint the image. + English image description (optional)

The Kontext model itself has a certain degree of image outpainting capability, but it is unstable with respect to prompts and input samples, which may lead to unintended modifications in non-specified areas. Therefore, we constructed an image outpainting dataset for Kontext using the EliGen dataset and trained this LoRA model. It can stably perform image outpainting and also generalize well to image inpainting and local editing scenarios.

As shown below, given an image with missing areas, the model can directly perform inpainting to complete the missing parts.