Chinese Text Image Generation

Open-source model for accurate Chinese glyphs in generated and edited images.

LongCat-Image Try Online

Why Chinese text in images is hard

Most text-to-image models struggle with Chinese characters: stroke order, rare glyphs, and layout in posters or UI mockups often produce garbled or missing text. Developers searching for a Chinese text image generation model need both generation quality and glyph accuracy.

LongCat-Image capabilities

LongCat-Image is a 6B open-source model optimized for Chinese text rendering and editing:

  • ChineseWord: 90.7 — strong coverage of standard characters
  • 8,105 standard Chinese characters including rare characters and calligraphy styles
  • Text-to-image and 15 editing tasks via natural language
  • Multi-round editing without quality degradation
  • Open-source SOTA on GEdit-Bench and ImgEdit-Bench for editing
GitHub Hugging Face

Use cases

  • Marketing posters and banners with embedded Chinese copy
  • UI/UX mockups with localized text
  • Educational materials requiring precise character forms
  • Iterative image editing: change text, style, or layout in follow-up rounds

Also available on LongCat Web and the LongCat APP — see Ecosystem.

Related