ByteDance unveils new AI image model to rival Google DeepMind's 'Nano Banana'
1–7 of 7 posts
Re: ByteDance unveils new AI image model to rival Google DeepMind's 'Nano Banana'
#2Re: ByteDance unveils new AI image model to rival Google DeepMind's 'Nano Banana'
#3Anything like nano banana in terms of quality that can run locally?
[1] https://docs.comfy.org/tutorials/image/qwen/qwen-image-edit
[2] https://docs.comfy.org/tutorials/flux/flux-1-kontext-dev
Re: ByteDance unveils new AI image model to rival Google DeepMind's 'Nano Banana'
#4https://genai-showdown.specr.net/image-editing
An important callout is that it is able to handle significantly larger native resolutions without having to downscale before making edits. Unlike many other image-to-image GenAI models, which cap out around 1 MP, Seedream can handle images up to 4K resolution, and this means less overall image degradation.
It also has pretty decent text-to-image ability as well though gpt-image-1 is still the heavyweight in that category when strictly considering prompt adherence.
Re: ByteDance unveils new AI image model to rival Google DeepMind's 'Nano Banana'
#5Re: ByteDance unveils new AI image model to rival Google DeepMind's 'Nano Banana'
#6Anything like nano banana in terms of quality that can run locally?
Nano Banana is still at the top of the class but there are two fairly decent image editor style models that you can run locally - Qwen Edit 20b [1] and Kontext Dev [2]. [1] https://docs.comfy.org/tutorials/image/qwen/qwen-image-edit [2] https://docs.comfy.org/tutorials/flux/flux-1-kontext-dev