ByteDance Releases Seedream 5.5 With Support for Up to 100 Editing Rounds

According to ChainCatcher, ByteDance’s Seed team released its new image-generation model, Seedream 5.5, on Nov. 9. The model can support up to 100 rounds of continuous editing on one image while preserving unedited areas at the pixel level, reducing color shifts and limiting detail loss and image-quality degradation. It also natively generates transparent or semi-transparent images with an Alpha channel and improves rendering of lighting, materials, skin textures and text. Evaluations found higher generation scores and usability rates than Seedream 5.0 Pro across six areas: e-commerce, marketing, film and television, gaming, office work and fashion design. Usability rates in film and office scenarios were more than double those of the previous version. In text-structure and realism tests covering text-to-image and image-to-image tasks, all four comparisons with the previous generation showed a positive preference; the realism image-to-image GSB score was +55.62%. Seedream 5.5 also outperformed GPT Image 2.5 Sunburst and Nano Banana 2.1 in text structure and text-to-image realism. In tests covering Chinese and English text, including dense small text, severe text-rendering failures fell to 0.4% and minor failures to 1.7%. The model is rolling out on Jimeng, Doubao, Doubao Work and Volcano Ark. A Pro version is planned.