Gemini 2.0 Flash (Image Generation) Experimental is Not working properly

Gemini API
, ,
1

I am trying to implement this model in my application but it always blocks responses without any reason even I am not providing harmfull or sexual images to it for re-generations. Google please fix this as GPT-4o image generation capabilities already touching sky.

3

Hi Suparva,

Sorry for the late response. Could you please try with the Gemini 2.0 Flash Preview Image Generation model and let us know if you are still experiencing the blockage in the response?

5

Hi @Suparva ,
Gemini 2.0 Flash (Image Generation - Experimental) is blocking safe prompts without any clear reason, even with safetySettings set to BLOCK_NONE. This makes it hard to use in production. Please clarify the limitations or fix the filtering logic. GPT-4o image generation is far more consistent right now.

Thanks.

6

I attempted to create multiple Ghibli-style images, but they were consistently blocked for unspecified reasons. Even though I understand the safety limitations and provided similar images to GPT-4o, it functioned successfully. My primary concern is the time constraint for publishing new content, while Google is addressing its issues. Consequently, I’ve now switched to OpenAI for image generation, even at a higher cost. The delay has become significant, and any potential Google fixes are now irrelevant to my current needs.