I wonder what model is Gemini 2.0 Flash Image Generation using…
Anyway, it’s pretty frustrating and difficult to generate images because I keep getting hit with the ‘Content not permitted’ restriction, even though I don’t understand what unusual requests I’m making (I send a picture and ask to remove extra text in it, or tell a story with a character in the image, or add something to my picture, etc.). I’ve even turned off all the Safety Settings.
benniepie , No, it’s a native Gemini 2.0 capability. From what I understand, it generates images in base64 encoding.
Rin_Rin , ‘Content not permitted’ is a persistent issue since I started trying Gemini 1.5. In short, if you look in the chat file in your drive/AI studio folder (download it, open it with Notepad++/VS Code or anything that can handle JSON schema) and scroll to the message that was “not permitted”, there will be a “finishReason” parameter with some value. In this particular scenario, it will likely be “IMAGE_SAFETY”. So… Google just slapped your wrist for trying to generate something NSFW (or at least it’s a false positive). I think they implemented a safety filter and don’t tell anyone about it, and furthermore, they do not provide a slider to adjust it to your settings. Kinda a shady move imo.
really do not understand why google launches these products proudly and the image edit only works for few in the market, and not for all, with the so called content restricted policy etc., weird policies. image edit does not work.