---
title: MiniT2I
sdk: static
app_file: index.html
pinned: false
---

# MiniT2I

Introducing **MiniT2I**: a simple pixel-space text-to-image generator baseline. Please use the project links below to navigate.

## Links

- Blog: [Text-to-Image Generation Made Simple](https://peppaking8.github.io/#/post/minit2i)
- PyTorch/Diffusers code: [Hope7Happiness/t2i-release](https://github.com/Hope7Happiness/t2i-release)
- JAX code: [PeppaKing8/minit2i-jax](https://github.com/PeppaKing8/minit2i-jax)
- PyTorch checkpoints: [MiniT2I/MiniT2I](https://huggingface.co/MiniT2I/MiniT2I)
- JAX checkpoints: [MiniT2I-B/16](https://huggingface.co/MiniT2I/MiniT2I-B-16-jax), [MiniT2I-L/16](https://huggingface.co/MiniT2I/MiniT2I-L-16-jax), [MiniT2I-B/16-MF](https://huggingface.co/MiniT2I/MiniT2I-B-16-MF-jax)

## Models

| Model | Params | Patch | GenEval | DPG-Bench |
| --- | ---: | ---: | ---: | ---: |
| MiniT2I-B/16 | 258M + 341M text encoder | 16 | 0.874 | 84.6 |
| MiniT2I-L/16 | 912M + 341M text encoder | 16 | 0.887 | 85.5 |

## Citation

```bibtex
@misc{minit2i2026,
  title  = {MiniT2I: Text-to-Image Generation Made Simple},
  author = {Wang, Xianbang and Zhao, Hanhong and Lu, Yiyang and Zhou, Kangyang and Ma, Linrui},
  year   = {2026},
  url    = {https://peppaking8.github.io/#/post/text-to-image-generation-made-simple}
}
```