Greetings. Am new to this; trying to convince myself not to hate it.

Can this image be altered (or create a new one) rotating the tracks 90 degrees so the tracks go on a slope from land to river entering underwater and mount the turbine on wheels, also rotated 90 degrees so the river flow can spin it ?

English is not my language and perhaps cannot express properly to GPT to create the image as desired. As far as I got :

Changes desired :

• 'rollercoaster' type tubular rails/tracks instead :

• turbine mounted on rollercoaster 'wheels' :

• sloping from ground to river.

• Thank you for your patience...

Try it in Gemini instead. Easier to refine the first few drafts with followup prompts

Thank you ! 👏 Your suggestion yielded this; need to turn the turbine 90 degrees... 😀

Tell it to “align” the turbine with the flow of the water

OK. It resulted very, VERY close to as intended.

Will try "align" instead of the failed 'turn 90 degrees' or 'transversal'.

Edited November 28Nov 28 by Externet

Getting there... Wrong angle to the stream flow; changed itself to different wrong wheels...

--> Question... Who owns the image copyright ? - - - The user of A.I. or Google ? 😳

Interesting river under river... 😄

30 minutes ago, Externet said:

--> Question... Who owns the image copyright ? - - - The user of A.I. or Google ? 😳

The photographer that took the photo that the model was trained on?

More seriously; difficult to provide a generally applicable answer. It may differ between regions, product lines, license etc. Short answer, for a few services I am familiar with, you own the output the AI generates. Issues may involve if the AI outputs very similar images for several users or if output includes other copyrighted material.

16 hours ago, iNow said:

Tell it to “align”

That is a good suggestion since the model may have limitations of analysing its own output. A prompt "rotate 90 degrees" may cause issues if the model does not manage to grasp what to use as a reference.

@Externet this is a case where prompting have limited success; there may not be many similar constructions in the training data. If the AI tool allows you may experiment with uploading references or modifying the image in an image editor and then submit it to AI for "cleaning".

Example: I rotated the turbine manually in an image editor. No need to be precise; there are artefacts.

Submit the above image to AI, prompting something like "just clean up this picture a little bit"

Result:

(If more control is required there are the concepts of control nets and similar, but that may require local installations, configurations etc. I'll can provide some links if interesting.)

Note @Externet from the title I got the impression you wanted to use ChatGPT; the cleaned up image above was created using ChatGPT, model GPT 5.1

Edited November 28Nov 28 by Ghideon
moved side note to the end

2 hours ago, Ghideon said:

Example: I rotated the turbine manually in an image editor. No need to be precise; there are artefacts.

Thank you, Ghideon. You did better than A.I. 😀 That is it. Perhaps I will make it appear bigger, like 10 metres diameter turbine, if I find the right words/commands...

1 hour ago, Externet said:

Perhaps I will make it appear bigger

Resizing the turbine after rotating, before submitting it to AI? Example; not 10 meters but bigger that before: