Hello,
I’ve been using ChatGPT’s image generation feature for a group portrait project that requires accurate positioning and character consistency. However, many generations failed (missing characters, incorrect positions or styles), yet were still counted against my daily quota.
Also, memory across sessions often resets, forcing me to repeat the same instructions, which increases time and frustration.
Suggestions:
1. Allow marking failed generations as invalid, exempting them from quota.
2. Strengthen conversation memory across sessions.
3. Add a “Report failed image” option for quota protection.
Thank you for your attention.