ShareGPT-4o-Image: Aligning Multimodal Models with GPT-4o-Level Image Generation

TL;DR

This paper introduces ShareGPT-4o-Image, a large synthetic dataset, and Janus-4o, a multimodal model capable of high-quality text-to-image and text-and-image-to-image generation, advancing open research in photorealistic image synthesis.

Contribution

The paper presents the first synthetic dataset for multimodal image generation and a new model, Janus-4o, that improves image generation quality and supports new tasks using limited training data.

Findings

Janus-4o outperforms previous models in text-to-image generation.

Janus-4o successfully performs text-and-image-to-image generation from scratch.

The approach achieves high-quality results with only 91K synthetic samples and 6 hours of training.

Abstract

Recent advances in multimodal generative models have unlocked photorealistic, instruction-aligned image generation, yet leading systems like GPT-4o-Image remain proprietary and inaccessible. To democratize these capabilities, we present ShareGPT-4o-Image, the first dataset comprising 45K text-to-image and 46K text-and-image-to-image data, all synthesized using GPT-4o's image generation capabilities for distilling its advanced image generation abilities. Leveraging this dataset, we develop Janus-4o, a multimodal large language model capable of both text-to-image and text-and-image-to-image generation. Janus-4o not only significantly improves text-to-image generation over its predecessor, Janus-Pro, but also newly supports text-and-image-to-image generation. Notably, it achieves impressive performance in text-and-image-to-image generation from scratch, using only 91K synthetic samples and 6 hours of training on an 8 A800-GPU machine. We hope the release of ShareGPT-4o-Image and Janus-4o will foster open research in photorealistic, instruction-aligned image generation.