Scheduled Tasks use DALL·E 3 instead of GPT-4o for image generation — is there a way to fix this?

ChatGPT Feature requests
,
1

Hi!

I’ve encountered a limitation with image generation when using GPT-4.0 with Scheduled Tasks Beta, and I’d like to clarify whether there’s a workaround — or at least whether improvements are planned.

For context: I’m using a paid Plus subscription, so all of this is happening inside the premium GPT-4 experience, not the free tier.

Here’s what’s happening:

  • All interactions take place within GPT-4.0 with Scheduled Tasks Beta.
  • When I schedule a task that generates an image for later execution, the image is rendered using the older DALL·E 3 model — recognizable by its more limited detail, stiffness in style, and less contextual awareness.
  • However, if I ask the same model (GPT-4.0 Scheduled Tasks) to generate the same image in real time, it uses the new GPT-4o image generation system, which produces far better results in terms of visual quality and prompt adherence.

So essentially, the same GPT-4.0 environment is using two different engines for image generation depending on timing:

  • DALL·E 3 when the task is scheduled
  • GPT-4o (or GPT-4 with vision capabilities) when the task is executed immediately

There doesn’t seem to be a way to explicitly choose the engine when scheduling.

My questions:

  1. Is there any current method to force Scheduled Tasks to use GPT-4o for image generation?
  2. If not, is there a plan to enable this in the future? It’s quite important for workflows that rely on scheduled, high-quality visual content without compromising output quality.

Thanks in advance for any insight or updates!