ByteDance launches Seedream 5.5 image model with 100-round editing and far fewer text errors
ByteDance's Seed team has released Seedream 5.5, a new image creation model. It brings three key upgrades: better multi-round editing, native transparent image support, and more natural textures with clearer text. The model is rolling out on platforms like JiMeng, Doubao, and Volcano Ark.
For editing, the model keeps untouched areas pixel-perfect, allowing up to 100 rounds of changes on the same artwork while keeping the subject and layout consistent. It can also generate images with transparent backgrounds directly, so no manual cutouts are needed. For text rendering, the severe error rate dropped to 0.4%, and the minor error rate to 1.7%.
In tests across six scenarios such as e-commerce, film, and office work, the model scored higher than the previous version, with double the usability in film and office tasks. It also leads mainstream rivals in text accuracy and realism. A 2K image edit costs about 0.28 yuan, less than half of the older version. A Pro version is coming later.
Turn text prompts and reference images into polished, professional visuals for work, social media, and presentations—no design skills required.
ByteDance AI chatbot development platform with LLM integration, plugin ecosystem and cross-platform deployment to Doubao and Feishu
Comments
Loading comments…