xianz mainly has these situations. First, the quality of the image generated by the same prompt is very different. This may be an API issue; 2. After entering instructions, the image cannot be displayed, but it shows that tokens have already been used; 3. Sometimes it cannot generate an image. After giving it the next hint, it will combine the two prompts to generate a single image. The model tried 2.0 flash, 2.0 flash exp, etc.

Is mrguu using the chatboxai service? The first and third feeling is that the model itself is unstable, and second, an error has occurred, or is there no error report or no content?