Dear OpenAI Team,
I appreciate the incredible work you’ve done with DALL·E, but I’ve noticed a major issue that affects its usability in religious, technical, and scientific illustrations—DALL·E struggles with counting accuracy.
For example, when generating an image of Lord Surya (the Sun God) on his chariot, which is traditionally drawn by exactly seven horses, DALL·E frequently generates 6, 8, 9, or even 11 horses instead of 7. This inaccuracy creates problems in religious and culturally significant artwork, as well as technical illustrations that require precise numbers of objects.
Why is this improvement important?
Religious & Cultural Accuracy: Many mythological and spiritual illustrations require exact details.
Technical & Scientific Diagrams: Fields like engineering, biology, and mathematics demand precise object counts.
Educational Content: Incorrect figures can mislead learners and affect educational credibility.
Suggested Improvement:
Can DALL·E be trained to recognize and accurately generate a specific number of objects when explicitly mentioned in the prompt? A built-in counting mechanism or post-processing AI check could ensure numerical precision.
This enhancement would make DALL·E significantly more reliable for professional, religious, and scientific use cases.
Would love to hear OpenAI’s thoughts on this! Is this a known limitation, and is there any roadmap for improving counting accuracy?
Thanks for your time and for continuously improving AI-generated art! 
Best Wishes,
Shashank Tyagi.