ChatGPT's Developer Mode:
One-Word Image Commands
Upload a photo. Type one word. Get a cutaway, a blueprint, an x-ray, an exploded view. No prompt engineering required — just the right trigger word.
This is the first Developer Mode post — the corner of Editor's Notes for the unglamorous tricks that just work. No funnel, no offer, no "secret." Just a genuinely useful pattern for getting more out of a tool you're probably already using.
What's actually going on here
These aren't secret ChatGPT commands or hidden settings. They're one-word prompts — labels like /cutaway or /blueprint — that work because they're specific and unambiguous. "Show me the inside of this" is vague and the model has to guess what you mean. "/xray" tells it exactly what kind of reveal to attempt. The trick isn't a jailbreak. It's just precise prompting compressed into a single word.
Bait: "279 SECRET codes ChatGPT doesn't want you to know." These are custom labels someone wrote, not official commands, and there's nothing secret about them.
Real: the underlying technique works. A tight, specific one-word prompt genuinely produces cleaner results than a vague sentence — that part holds up regardless of the marketing around it.
How to use them
- 1Upload a photo, sketch, screenshot, product, person, or animal.
- 2Type only the codeword — nothing else. Example: just send
/cutaway. - 3If you need labels or an educational breakdown, say so in a short follow-up rather than the first message.
- 4To keep a real face, pose, or exact proportions intact, add one short line asking it to preserve those — some transformations will drift otherwise.
The starter kit
Instead of all 279, here are the ones worth actually keeping on hand — grouped by what you're trying to do.
/cutaway
Slices away part of the exterior to show what's inside.
/xray
Reveals internal structure straight through the surface.
/explodedview
Spreads every component apart while keeping positions clear — great for anything with visible parts.
/layers
Separates the subject into stacked layers, front to back.
/blueprint
Turns the photo into a technical blue-line drawing.
/schematic
Simplifies the subject into a functional diagram.
/isometric
Angled engineering-style view, useful for product or furniture shots.
/howitworks
Visually explains the basic function of whatever's in frame.
/beforeafter
Shows a change, repair, or transformation side by side.
/proscons
Lays out advantages and disadvantages clearly.
/scalecompare
Places the object beside familiar items or a person for intuitive size.
/stepsequence
Breaks a task into frame-by-frame steps — handy for how-to content.
/infographic
Explains the subject with clean visual information blocks.
/whiteboard
Turns the image into a hand-drawn whiteboard-style explanation.
/isometric
Also doubles as a clean product-shot style for social posts.
/annotatedphoto
Adds clean, direct labels onto the photo itself.
The trick isn't the word. It's that a one-word command removes every chance for the model to misread a vague sentence.
Who this is actually useful for
Content creators explaining a product. Sellers who need a clean listing image without a design tool. Anyone building a how-to post, a repair guide, or a comparison graphic. It's not going to replace real product photography or a CAD file, but for a fast, free visual explainer, it's a legitimately good shortcut — which is rarer than it sounds in this corner of the internet.
If you want the complete 279-entry codebook — including scientific views like thermal and infrared, worldbuilding styles like cyberpunk and steampunk, and repair/maintenance-specific codes — it's worth going straight to the source.
John Savage's Daily AI Newsletter
Where the full 279-code list originated — worth a subscribe if this kind of practical AI trick is useful to you.