---
tags:
- text-to-image
language:
- en
pipeline_tag: text-to-image
---

### Model

<img src="https://cdn-uploads.huggingface.co/production/uploads/629cf0475a13ba8233dd18c9/c0M8mGR5TlKJIsoHMallb.png" alt="drawing" width="400"/>

[Switti](https://arxiv.org/abs/2412.01819) - *S*cale-*wi*se *t*ransformer for *t*ext-to-*i*mage synthesis is a scale-wise transformer for text-to-image generation. 

### Model Description

- **Developed by:** Yandex Research
- **Model type:** Scale-wise text-to-image generative model

### Evaluation

<img src="https://cdn-uploads.huggingface.co/production/uploads/629cf0475a13ba8233dd18c9/TYUHk2DpFfGn2XMwD5QPR.png" alt="drawing" width="400"/>

The bar chart above shows user preference for Switti vs competitor models with respect to different aspects.