Simply the quality of image results is a far away from the chatgpt results. its not even possible to get a result on any named buildings i.e. Buckingham Palace etc. all photos, however prompted look like cheap midjourney scifi pics. in Chatgpt, the same prompt delivers extraordinary good results! very frustrating company policy and really preventing me from a real AI Pipeline!
As professionals using the DALL·E API, we’re handed the raw engine — no tires, no dashboard, no GPS. Meanwhile, ChatGPT users get beautifully optimized generations with minimal effort thanks to automatic prompt tuning, context retention, and orchestration layers we can’t access.Faces are left blank, Winston churchill looks like anything, but not him. buildings/architecture all blocked.. A useless tool like this.
We’re expected to reverse-engineer prompts, handle visual composition manually, and still end up with outputs that don’t match the polish of what’s available in ChatGPT.
We need more parity, transparency, or tools to bridge this gap — because right now, the “pro” experience feels ironically underpowered.