I’ve been experimenting with Gemini 2.0 Flash Experimental’s image editing capabilities, and I’m consistently running into a frustrating issue: the “content is not permitted” error.
I can confidently say that my inputs do not violate any policies or ethical guidelines.
What’s particularly puzzling is that this error seems to happen almost exclusively when I’m editing Japanese anime-style images. I’ve tried various prompts and tweaks, but the issue persists.
This severely impacts the usability of the feature, especially for those of us who enjoy working with this type of artwork.
Has anyone else encountered this issue? Any insights or potential workarounds would be greatly appreciated.
I don’t think it is limited to anime style image editing. I cannot edit any of my familiy pictures - totally harmless familiy pictures. (where noone is japanese )
Definitely the new image safety filter is either producing false positives and/or trained in a way that prohibits use of real-life faces (and any nsfw). From what I can tell, Gemini 2.0 img gen also struggles with bringing the character to life (eg. making them in realistic/photo style).
Read more about false positives and the new safety filter from my post here
I am not able to generate even a simple image. Keeps restricting. I put Cursor in an agent mode and it finally gave up. Check the screenshots. I wanted to try out flash 2.0 exp after reading so many good reviews on twitter only to be extremely disappointed again by Google again.
so far every image that involves a child gets blocked for me. i asked for images of individuals of a family- one at a time- from back view. I attempted child last and no dice. no issues at all with mom , dad, or aunt. prefaced that it was safe for all audiences and that was for children’s book illustration. still nothing. this needs to change. it is far from kiddie “Modified by moderator”. maybe if i give the kid a moustache, it will allow to create images.
First Reply on here since corse correcting on my end as the user … so please bare with me!
As a general rule of thumb … typically the longer you have your account the more adapt the AI/ML will be so your AI/ML typically learn over time… therefore if the account is yours you shouldn’t have an issue it should start picking up on the nuances and intricite recurring themes such as voice control…if you you use that accounts for a cross verification such as utilizing your Google to sign into Samsung or your iCloud to sign into your Spotify oftentimes the different systems or code will not work properly it takes time for it to course correct and all you have to do is go through every single setting line by line as if you were coding until it starts to pick it up
I suggest sandhave recommended to clients that since you are using Gemini 2.0 flash experimental …to submit the feedback under feedback this way at least it flags the teams the engineering teams then they can go on the back end and code itif you arent a code guru like some of yall…
If your particular account is part of shared work then it should still have a separate login and you can add someone with their credentials there is no reason for someone to sit there and utilize your account that is a violation of privacy but also of our general terms and agreements which everyone should have read if you are using an experimental beta model…
Since we’re all in this together I would also like to add and remind everyone that reverse engineering Trojan horsing slowing down a service or attacking the network in which you are utilizing to try to get work done or to try and to experiment with an experimental app is in fact illegal so again I would definitely brush up on all the legal especially the bylaws within your region because it may vary by region depending on where their legal systems are and where they are with compliance this is what I do on normal as part of my start up side hustle.