LongCat T2I

LongCat is Meituan's lightweight bilingual text to image and editing model, developed by the LongCat team and distributed as both open weights and a hosted API on fal.ai. It creates images directly from natural language descriptions with a focus on photorealism, coherent composition, and accurate placement of visual elements. The model is particularly known for rendering readable, well integrated text inside images, including multilingual text that follows perspective, lighting, and surface context. LongCat uses a compact 6 billion parameter architecture with multi aspect ratio bucketing, which keeps inference efficient while still supporting output up to roughly 4 megapixels. It is a strong choice for poster work, branded graphics, and creative production workflows that benefit from clean in image typography.

December 5, 2025

Fal.ai

Text to Image (T2I)