Dreamina 3.1 T2I

Dreamina 3.1 is ByteDance's text to image model focused on translating natural language prompts into images with stronger aesthetics and detail. It improves on earlier versions by enhancing style accuracy, richness of textures, and visual coherence. Dreamina 3.1 supports both English and Chinese prompts and is optimized for fast synchronous inference, with users able to request images at resolutions up to about 2048 pixels on the long edge, equivalent to roughly 4 megapixels. It is particularly well suited for creative projects where stylistic control and cinematic refinement matter, including design, editorial media, illustration, and marketing visuals. The model is exposed through API endpoints on platforms such as Replicate and WaveSpeed and is a solid default for polished brand imagery in multiple languages.

August 7, 2025

Bytedance

Text to Image (T2I)