ByteDance has released Seedream 5.5, a new AI image model that the company says can edit the same picture 100 times in a row without the image falling apart. It is rolling out slowly in Doubao, Jimeng and Volcengine, and ByteDance says it keeps unchanged parts of an image identical.
ByteDance has released Seedream 5.5, a new AI image model that the company says can edit the same picture 100 times in a row without the image falling apart.
Where Is Seedream 5.5 Available?
The Seed research team announced the model on Monday. It is rolling out slowly in ByteDance's Chinese apps, including Doubao and Jimeng, and on its Volcengine cloud platform.
What Can ByteDance's Seedream 5.5 Do?
Image editors often wear a picture down over repeated changes, because each output becomes the input for the next edit. ByteDance says Seedream 5.5 keeps the parts of an image that should not change identical, down to the pixel, with no color shifts. In one demo, the company edited the same portrait 100 times, swapping clothes and objects while still adjusting small items such as rings.
The model can also generate images with transparent backgrounds. Designers can place elements, such as a snowy owl, directly into layouts without cutting them out by hand. ByteDance also says the model renders skin, light and dense small text more realistically.
How Does Seedream 5.5 Compare With Rival Image Models?
In ByteDance's own tests, Seedream 5.5 beat OpenAI's GPT Image 2.5 and Google's Nano Banana 2.1 on text and realism. Those results have not been checked independently, according to The Decoder. ByteDance says serious text errors appeared in 0.4% of its test images.
Seedream sits alongside ByteDance's Seedance video model. Both are covered by a copyright agreement ByteDance signed with Hollywood's studios in August. The company said a Pro version of Seedream 5.5 is coming.
Este artigo foi escrito com a assistência de IA.
News Factory APP - notícias agênticas para impulsionar seu SEO e AEO.