Glyph-ByT5-v2 is an open source text coder from Microsoft

Upgrade to V2 version
Compared with the previous version focusing on English texts
Glyph-ByT5-v2 can support accurate spelling in 10 different languages, significantly improving the accuracy and breadth of multilingual text rendering.

The following content is from the original text:

The latest step-by-step perceptual preference learning (SPO) method is adopted to significantly improve the quality of visual aesthetics and make the generated images more visually attractive.
Recently, Glyph-ByT5 has achieved high-precision visual text rendering performance in graphic design images, but it still only focuses on English and performs relatively poorly in terms of visual appeal. In this work, we address these two basic limitations by proposing Glyph-ByT5-v2, which not only supports accurate visual text rendering in 10 different languages, but also achieves better aesthetic quality.

To achieve this goal, we have made the following contributions: (i) Create a quality multilingual glyph text and graphic design dataset that contains more than 1 million glyph text pairs and 10 million graphic design image text pairs, covering nine other languages,(ii) build a multilingual visual paragraph benchmark of 0 prompts, with 100 prompts per language, to assess multilingual visual spelling accuracy, and (iii) leveraging the latest progressive perceptual preference learning methods to improve visual aesthetic quality.

Through the combination of these technologies, we provide Glyph-ByT5-v2, a powerful custom multilingual text encoder, and Glyph-SDXL-v2, a powerful aesthetic graphics generation model, that supports accurate spelling in 10 different languages. Considering that the latest DALLE-3 and Ideogram are still handling multilingual visual text rendering tasks, we believe our work is a major improvement.

For more details, you can browse the link below the video
Thank you for watching this video. If you like it, please subscribe and like it. thank

Project address:https://glyph-byt5-v2.github.io
Model download:https://huggingface.co/GlyphByT5/Glyph-SDXL-v2

Oil tubing:

Scroll to Top