LongCat-Image-AIO

Overview

This repository contains the All-In-One (AIO) merged model for LongCat-Image, combining both Text-to-Image Generation and Image Editing capabilities into a single unified model. LongCat-Image-AIO merges the strengths of LongCat-Image-Edit and LongCat-Image-Edit-Turbo into one model — enabling both high-quality image editing and text-to-image generation without switching models.

Base Model: meituan-longcat/LongCat-Image
Merged Models:
- LongCatEdit-Full (Standard Editing)
- LongCatEdit-Light (Fast Editing)
Architecture: LongCat-Image (6B Parameters)
License: Apache 2.0

Key Capabilities

🎨 Text-to-Image Generation: Generate high-quality photorealistic images from text prompts with exceptional Chinese and English text rendering.
✏️ Image Editing: Perform precise instruction-based image editing — global editing, local editing, text modification, and reference-guided editing.
⚡ Turbo Editing (Fast Mode): Achieve high-quality image editing with only 8 NFEs (Number of Function Evaluations) for extremely low inference latency.
🌐 Bilingual Support: Full Chinese-English bilingual support with superior Chinese text rendering accuracy and coverage.
🔒 Consistency Preservation: Strong preservation of non-edited regions — layout, texture, color tone, and subject identity remain invariant unless targeted by the instruction.
📝 Specialized Text Rendering: Utilizes a character-level encoding strategy for quoted content to ensure accurate text generation in images.

Text-to-Image Generation

📝 Special Handling for Text Rendering

For tasks involving text generation, you must enclose the target text within single or double quotation marks (both English '...' / "..." and Chinese '...' / "..." styles are supported).