leeguandong/Awesome-Chinese-Stable-Diffusion

What it solves

Finding tools for generating or editing images in Chinese is scattered across many repositories, papers, and platforms. Most image-generation resources are built for English, and Chinese-specific models, evaluation benchmarks, or datasets can be hard to locate. This project collects and organizes those resources in one place, making it easier for researchers, developers, and artists to discover what exists and how it compares.

How it works

This is a manually curated list, not a software tool. The README groups content into three main sections: models, evaluation, and datasets. The model section distinguishes open- and closed-source options, lists recent foundation-model progress, and provides a summary table with parameters, architecture type, text encoder, resolution, and whether they support rendering Chinese text. Each model entry links to its official repository or model card and includes a short description. Evaluation resources cover benchmarks, tools, and leaderboards, while dataset resources list training and evaluation/annotation sets. The project invites contributions via pull requests, asking for repository links, star counts, and brief descriptions in the same format.

Who it's for

This is aimed at anyone working with Chinese-language image generation or editing: researchers comparing models, developers choosing a model for an application, hobbyists exploring the field, or maintainers tracking the latest Chinese-specific resources. It is also useful for people who want quick access to datasets or evaluation tools tailored to Chinese text-to-image tasks.

Highlights

• Covers both open and closed models, from early Stable Diffusion variants to recent DiT, autoregressive, MoE, and unified multimodal architectures
• Includes a comparison table with key specs (parameters, architecture, text encoder, resolution, Chinese text rendering support)
• Links each model to its official source (GitHub, ModelScope, Hugging Face, or project page)
• Also lists evaluation benchmarks, evaluation tools, leaderboards, training datasets, and annotation datasets
• Actively invites community contributions via pull requests, with a clear format for adding new entries