Xiaohongshu Releases FireRed-Image-Edit 1.1 Model for Advanced Image Editing

The team at Xiaohongshu has released FireRed-Image-Edit 1.1, an upgraded image-editing model designed to improve identity consistency, multi-element composition, portrait beauty editing, and font-style referencing.

According to the company, the model introduces what it describes as state-of-the-art identity preservation technology in the open-source ecosystem, enabling edited images to maintain clear recognition of the original subject even after complex modifications.

The model can combine more than 10 different visual elements within a single image and includes an agent-based intelligent cropping and compositing feature that automatically arranges visual components.

For portrait editing, FireRed-Image-Edit 1.1 provides dozens of makeup and beauty styles, allowing users to perform professional-grade retouching and create customized looks such as brightening yellow or olive skin tones.

The system also features an open LoRA training ecosystem, enabling creators to develop and apply custom visual styles. In addition, the model includes an optimized sampler designed to maximize GPU efficiency when handling identical tasks, image sizes, and input volumes.

With 30GB of GPU memory, FireRed-Image-Edit 1.1 can generate edited images end-to-end in about 4.5 seconds. The model supports cross-platform deployment, offers native node compatibility with ComfyUI, and supports the GGUF lightweight format, helping accelerate training and inference workflows.

Source: IT Home