# Text-to-Image Generative Modeling Without Language Pretraining

> Source: <https://aiflash.com/news/135822/>
> Published: 2026-10-20 06:30:00+00:00

Modern text-to-image (T2I) systems use internet-scale language pretraining for text-conditioned image generation. We ask whether language pretraining is necessary for strong T2I generation. We introduce Halcyon, a family of early-fusion multimodal autoregressive models pretrained entirely from scrat