ByteDance unveils seedream 5.5, a new image generation model capable of extensive iterative editing and transparent output

ByteDance's Seed team has released Seedream 5.5, a next-generation image creation model featuring significant upgrades. Key enhancements include highly stable multi-round editing, allowing for up to 100 iterations while maintaining pixel-level fidelity in non-edited areas, and native RGBA transparent material generation, eliminating the need for post-production masking. The model also boasts improved realism in lighting, texture, and skin rendering, alongside a marked boost in text clarity. Testing across six scenarios—e-commerce, marketing, film, gaming, office, and fashion design—showed that Seedream 5.5 outperformed Seedream 5.0 Pro, with usability rates in the film and office sectors doubling. In specialized evaluations, Seedream 5.5 achieved a positive preference in four comparisons against the previous version. Its severe text corruption rate dropped to 0.4% across Chinese and English text, significantly better than the prior version. The model is now available on platforms like Jimeng and Doubao, with the API pricing for a 2K image edit being approximately 0.28 yuan, less than half the cost of Seedream 5.0 Pro. ByteDance plans to release a Pro version and expand its application within Agent workflows.