wind nexus
Hi everyone,
I’m trying to understand whether others are seeing similar behavior with ChatGPT image generation.
For several days I’ve been getting frequent failures with the message:
“We experienced an error when generating images.”
What makes this confusing is:
-
Sometimes generation works briefly
-
Then identical or even simpler prompts fail again
-
Even ultra-minimal prompts like:
- “red circle on white background”
- “minimal educational slide”
can fail intermittently
-
VPN changes do not seem to help
-
The issue appears both in browser and desktop app
-
Some retry/cycle attempts occasionally succeed temporarily
What I already tested:
- Different browsers
- VPN on/off and different regions
- New chats
- Very small prompts
- Waiting between retries
- Different image complexities
- Different sessions/devices
The strange part is that this does not behave like a normal policy refusal or account restriction.
It feels more like an unstable backend/pipeline issue, overloaded queue, routing problem, or failed worker allocation.
Questions:
- Is anyone else seeing intermittent image failures lately?
- Do certain models/routes seem more stable than others?
- Has anyone identified reliable mitigation strategies?
- Could this be related to account-level rate scoring, backend region routing, or image worker availability?
Would appreciate any technical insights or comparable experiences.
Thanks!