All-in-One Photo Editor for Portraits, Makeup, and Multi-Image Composition
Editing a face with most AI tools is a gamble — ask for a makeup change and the person often comes back looking like someone else. FireRed-Image-Edit-1.1 is built to keep the subject recognizably themselves through even dramatic edits, the way a skilled retoucher can restyle a portrait without changing who is in it. It rolls portrait retouching, makeup styling, multi-image blending, text-style transfer, and old-photo repair into one model, and ships with a speed-tuning kit so it is practical to run in real production. Built by the FireRedTeam.
What it does
FireRed-Image-Edit-1.1 is a general-purpose image editing foundation model developed by FireRedTeam. It is an upgrade over FireRed-Image-Edit-1.0, with significant enhancements in identity consistency, multi-image conditioning, and domain-specialized editing. The model is designed to bridge the gap between open-source capabilities and closed-source production solutions, excelling at portrait editing, multi-subject composition, makeup styling, text style reference, and photo restoration. It is built for real-world creative production workflows and comes with an extensive engineering optimization suite for deployment at scale.
Problem it solves
- Identity consistency – Open-source state-of-the-art in character identity preservation, keeping subjects recognizable across complex and imaginative edits
- Multi-element fusion – Freely combines 10+ input elements with Agent-powered automatic cropping and stitching, eliminating the need for lengthy prompt engineering
- Portrait & makeup editing – Dozens of makeup and beauty styles, from professional retouching to creative Halloween looks and skin tone enhancement
- Text style reference – Maintains high-fidelity typography and stylized text rendering, on par with closed-source solutions
- Photo restoration – High-quality old photo repair with superior fine-grained detail recovery
- Production deployment – Designed for real-world use with ComfyUI support, GGUF format, LoRA training, and extreme speed optimization
Input/Output
- Input: One or more images + a natural-language editing instruction
- Supports portrait editing, multi-image fusion, makeup styling, text-style reference, virtual try-on, photo restoration, and style transfer
- The Agent workflow handles complex multi-image compositions automatically
- Output: A high-quality RGB image with the edits applied
- If a reference image is provided, it is used as guidance (e.g. copy specific details from the reference)
- If no reference image is provided, the result is generated from the text prompt alone
- Parameters:
- Diffusion steps — 5–40 (higher = better quality, slower)
- Output image size — width × height, 512–1600
- Random seed (Advanced) — enter a number for reproducibility, or -1 to randomize
A sample workflow from community
Accuracy & Speed
Benchmark standing | Open-source SOTA on ImgEdit, GEdit, and RedEdit; surpasses closed-source competitors in specific dimensions |
Human evaluation | Rated highly for prompt following and visual consistency |
Generation speed | ~4.5 seconds end-to-end |
VRAM needed | 30GB with the full optimization suite enabled |
Acceleration | Distillation + quantization + static compilation |
Model Source
- HuggingFace:
Compliance & Provenance
Provider | Open-source |
Provider type | Specialized |
License | |
EU AI Act risk class | Limited Risk |
Art. 50 transparency | Required — outputs are marked. See AI Policy §2. |
Region availability | Available globally |
Training data summary | Pending — provider has not yet published per Art. 53(d) |
For more on how we classify models and mark outputs, see our AI Policy.