I am not able to get the right image for the kind of automation solution i describe, could someone help me with the prompt or with getting better results?
Maybe?
I would need you to be more detailed than what you’ve posted so far, however.
If you have a native language other than English, I would prompt in that, because the way that you’re wording things in English, it’s actually impossible for me to even imagine what exactly it is you want.
even asking the bot to assume what the most viable or feasible way to convert that image closer to your words makes it go a little crazy…
Try describing just one shelf and not the whole machine
and then build/edit from there…
Please share what you get from either a native prompt or just describing how a single shelf is set up without forcing the model to resolve the whole machine at once…
Here is what gpt-5.5 says about prompting in English:
For gpt-image-2, English is generally the safest choice, especially for complex, highly specific, or stylistically detailed prompts. Image models are typically trained with a large amount of English captioning and art-direction language, so English prompts may produce more reliable results for nuanced instructions.
Your native language can also work well, particularly for simple descriptions or when expressing culturally specific ideas. If the prompt includes idioms, regional references, or subtle visual details, results may vary depending on how well the model understands those terms.
In short: English is usually best for maximum consistency, while your native language is acceptable for straightforward prompts or culturally specific concepts.
i’m 80% certain the Ai thinks there are wheels in the boxed areas at each end of the tracks.
Generating specific mechanical designs is always tricky because DALL-E/GPT-image-2 tries to make things look ‘cool’ rather than ‘functional’.
A few tips to fix your prompt:
-
Focus on one module first: As @windysoliloquy suggested, try prompting just for a ‘single shelf segment with a motorized robotic platform’ instead of the whole rack.
-
Use ‘Cutaway view’ or ‘Technical Schematic’ style: This forces the model to focus on the alignment of wheels and tracks rather than cinematic lighting.
-
Be specific about the placement: Instead of ‘at extreme ends’, try: ‘A low-profile robotic shuttle on four wheels, positioned between two vertical steel rails. The shuttle is aligned to move depth-wise into the shelving unit.’
Also, if you have a rough sketch, try using the ‘Edit’ (In-painting) tool in ChatGPT to highlight just the wheel area and describe only that part. It’s often better than re-generating the whole image from scratch.
maybe, but my concept has wheels at the extreme end / corner of shelf
Use ‘Cutaway view’ or ‘Technical Schematic’ style- is that a setting?
and you should be explaining that to the ai, because now already you’ve located the wheels with a descriptor.
that’s what you have to do with each piece of your machine that you go into a ‘new innovative idea’ with… give it the right descriptors that do what it needs to do…
and one of those important things happened when you used words to position the ambiguous part of your prompt
you want precision?
spend the extra words to be very precise 