Topic Prompt Creator

1 · Choose a topic

2 · Choose media

3 · Mix and match

Machine learning

Start from a whole sample, then swap in fields from the others.

Subject

Who or what is the main focus? Describe how it looks.

Pick one

Action

What is the subject doing, or what is it showing?

Pick one

Location

Where is it? Include the setting or background.

Pick one

Composition

Framing, camera angle and where things sit in the frame.

Pick several

Style

Medium, art style, colour palette and mood.

Pick several

Refiner

Aspect ratio, labels, things to avoid and final touches.

Pick several

Your prompt

ImageMachine learning

6 of 6 fields filled570 characters

Subject: A small hiker carrying a lantern labelled 'model'
Action: Stepping carefully down a foggy slope towards the lowest point, each footprint a little shorter than the last.
Location: On a smooth bowl-shaped 3D loss surface with contour lines drawn on the ground.
Composition: Wide shot showing the whole valley, trail of footprints curving down to the minimum, lowest point marked with a small flag.
Style: Clean 3D render with soft shadows, cool blue and teal palette, calm, curious mood.
Refiner: 16:9, label only 'loss' and 'minimum', no other text, high detail.

Be concrete about the subject and where it sits in the frame. Image tools often garble equations and long labels, so keep any text short and check it in the result.

ChatGPT Gemini

Avoid conflicting choices. When you mix samples, check that the fields agree — for example one aspect ratio, and one set of characters. Contradictions can give odd results, or none at all.