#Dalle3 from custom gpt not generating a single image ... for paid users..
1 messages · Page 1 of 1 (latest)
why is openai support so slow?
Hi jr,
I understand how frustrating that must feel—especially when you've seen the feature work differently in other contexts. At the moment, when using a custom GPT, image generation requests with DALL·E 3 can sometimes return text instead of an image if the model's configuration or permissions aren't set up to enable image output, or if the system encounters a restriction that prevents fulfillment of the request in that environment.
Free-tier ChatGPT users may appear to have wider image generation in some cases because they are using the standard, first-party interface, which is preconfigured to allow multiple images per request. Custom GPTs, however, operate within the boundaries of their creator-defined tools and the account's current allowances—so if the image generation tool isn't explicitly enabled during setup, or if it's limited to a single image per request, results may default back to descriptive text responses.
Here are a few steps you could try to get it working:
Open your custom GPT's settings and confirm that the “image generation” tool is enabled.
Make sure your prompt directs the model clearly to create an image, not just describe one—e.g., start with “Create an image of…” rather than “Describe an image of….”
If you're expecting multiple images, check whether that's currently supported in your access tier for DALL·E 3. At present, not all environments or subscription levels allow multi-image output.
If everything appears to be configured correctly but you still get only text back, it's worth sending details to the support team—including the exact prompt, the model you selected, and the tool settings—so they can check if this is a known limitation or a bug specific to your account.
I can also walk you through a step-by-step setup for enabling image generation correctly in a custom GPT so you can confirm whether it's a settings issue or a platform limitation.