100 ChatGPT Image Shortcuts to Create Better AI Images (2026)

Most people make AI image prompts harder than they need to be.

You don't need a 200-word prompt to get a good image. A short command tells ChatGPT how the image should look. A few keywords tell it what should appear. That is the whole system.

Three examples:

Create image /xray electric vehicle battery pack

Create image /iconset SEO, paid ads, CRM, email, analytics

Create image /xray /labelled /blueprint electric vehicle battery pack

One thing to be clear about before you scroll. These are not built-in ChatGPT features. There is no hidden slash-command menu for image generation. They work because ChatGPT reads them as plain instructions in context, not because OpenAI shipped them as commands. Treat them as your own shorthand: the value is that you write less, not that ChatGPT unlocks something new.

The logic behind every prompt below

Four parts, and every prompt in this post is built from them.

• Command — how it should look.
• Keywords — what should appear.
• Reference — what it should match (optional)
• Create image — the instruction to actually generate something.

Separate those four and prompting gets faster and far more predictable. The full list of commands is below. The rules for using them come straight after.

The 100 image shortcut commands

Illustration and art styles (1–20)

1. /sticknotes — Turn the idea into a composition made from sticky notes with handwritten labels.

Create image /sticknotes content marketing strategy, SEO, social media, email, paid ads, CRM

2. /handwritten — Turn text and annotations into natural handwritten lettering.

3. /doodle — Render the subject as playful hand-drawn doodles.

4. /sketch — Convert the scene into a pencil or ink concept sketch.

5. /lineart — Render everything with clean minimal outlines.

6. /whiteboard — Explain the idea visually as if someone drew it on a whiteboard.

7. /chalkboard — Chalk illustration and handwritten annotations on a dark board.

8. /watercolor — Render the subject in soft watercolor washes.

9. /gouache — A matte painted illustration with bold, flat forms.

10. /oilpaint — Transform the scene into a traditional oil painting.

11. /paperart — Rebuild the visual with layered paper textures.

12. /papercut — A dimensional cut-paper composition.

13. /collage — Turn the visual into an editorial mixed-media collage.

14. /clay — Recreate the subject as a soft 3D clay model.

15. /felt — Render the subject using handmade felt textures.

16. /embroidered — Make the image look stitched onto fabric.

17. /lowpoly — Convert objects into low-poly geometric forms.

18. /pixelart — Transform the subject into chunky retro pixel art.

19. /voxel — Build the scene using small 3D voxel blocks.

20. /isometric — Convert the scene into an isometric illustration.

Explainer and information visuals (21–40)

21. /xray — Make the exterior transparent and reveal the internal parts.

Create image /xray smartphone, battery, motherboard, camera module, speakers, sensors

22. /cutaway — Remove part of the exterior to show what is happening inside.

23. /crosssection — Slice through an object or environment to show its internal structure.

24. /exploded — Separate each major component into an exploded technical view.

Create image /exploded mechanical watch, dial, hands, movement, gears, case, crown, strap

25. /layers — Separate the subject into stacked layers.

26. /anatomy — Break the subject into labelled anatomical components.

27. /blueprint — Turn the idea into a technical blueprint.

28. /schematic — Represent the system using simplified technical connections.

29. /wireframe — Render the subject as a 3D wireframe.

30. /diagram — Convert the idea into a clean explanatory diagram.

31. /flowchart — Turn a process into connected steps and decision points.

32. /timeline — Show information in chronological order.

33. /beforeafter — Show two states side by side.

34. /comparison — Compare two or more objects or concepts visually.

35. /process — Illustrate a workflow step by step.

36. /howitworks — An annotated visual explaining how something works.

37. /ingredients — Show all components surrounding the finished result.

38. /breakdown — Break a complex idea into simpler visual sections.

39. /mapview — Present information as a map with zones or routes.

40. /systemmap — Show relationships between different parts of a system.

Icons, symbols and UI assets (41–55)

41. /iconset — Generate a consistent family of icons.

Create image /iconset analytics, CRM, automation, email, calendar, database, security

42. /glyphs — Minimal monochrome glyph icons.

43. /outlineicons — Icons built from consistent thin outlines.

44. /filledicons — Simple solid icons.

45. /3dicons — Turn concepts into polished 3D icons.

46. /clayicons — Soft clay-style icons.

47. /glassicons — Translucent glass-style icons.

48. /stickerpack — Turn subjects into a consistent sticker collection.

49. /badges — Badge-style graphics for multiple concepts.

50. /emoji — Turn concepts into custom emoji-style graphics.

51. /appicon — Turn an idea into a polished mobile app icon.

52. /favicon — Reduce the concept into an extremely simple, recognisable symbol.

53. /pictogram — Universal pictogram-style symbols.

54. /mascotset — Multiple poses or expressions of one mascot.

55. /uiassets — A visually consistent set of UI graphics.

Photography and product visuals (56–70)

56. /productshot — A polished commercial product photograph.

57. /heroshot — A dramatic premium hero image built around one focal object.

58. /floatingproduct — Suspend the product elegantly in mid-air.

59. /levitate — Make multiple objects float dynamically around the subject.

60. /flatlay — Arrange objects neatly and shoot them from directly above.

61. /macro — An extreme close-up emphasising material and texture.

62. /studio — Place the subject inside a minimal photography studio.

63. /softbox — Soft, diffused commercial lighting.

64. /spotlight — Illuminate the subject with dramatic directional light.

65. /rimlight — Add strong edge lighting around the subject.

66. /backlit — Light the subject mainly from behind.

67. /silhouette — Reduce the subject to a strong silhouette.

68. /reflection — Add controlled reflections around or below the object.

69. /shadowplay — Use shadows as a major part of the composition.

70. /materialstudy — Focus on material, texture and surface detail.

Camera and composition (71–80)

71. /topdown — Frame the scene from directly above.

72. /birdsview — Show the subject from a high aerial perspective.

73. /wormsview — Shoot dramatically upward from near ground level.

74. /closeup — Frame tightly around an important detail.

75. /ultrawide — Use a wide environmental composition.

76. /symmetry — Build the entire composition around visual symmetry.

77. /centerframe — Position the subject exactly in the centre.

78. /negative — Leave generous negative space around the subject.

79. /ruleofthirds — Place the focal subject using rule-of-thirds composition.

80. /depth — Create strong foreground, midground and background separation.

Typography and annotation (81–90)

81. /typographic — Make typography the main visual element.

82. /lettering — Create expressive custom lettering.

83. /scribbletext — Use loose, scribbled handwriting.

84. /markertext — Thick marker-style lettering.

85. /labelled — Add clean labels pointing to important parts.

86. /annotation — Add handwritten editorial notes and arrows.

87. /circled — Highlight elements using hand-drawn circles.

88. /highlighter — Marker-style highlighting over important details.

89. /redacted — A document-style visual with selected information blacked out.

90. /newspaper — Transform the composition into a newspaper-style editorial layout.

Creative transformation (91–100)

91. /miniature — Rebuild the subject as a tiny miniature world.

Create image /miniature construction site, tower crane, excavator, workers, scaffolding, cement mixer

92. /diorama — Place the entire scene inside a handcrafted diorama.

93. /toy — Turn the subject into a collectible toy.

94. /actionfigure — Turn a person or character into a packaged action figure.

95. /packaging — Place the subject inside premium retail packaging.

96. /museum — Display the object like an artefact inside a museum.

97. /billboard — Place the creative realistically onto an outdoor billboard.

98. /magazine — Convert the image into a premium editorial magazine cover.

99. /poster — Recompose the idea as a campaign poster.

100. /infographic — Convert information into a structured visual infographic.

10 rules that make these shortcuts work

The commands above are only half of it. These rules are what separate a usable image from a re-roll.

1. Always start with “Create image”

Give a direct generation instruction.

Do: Create image /xray electric vehicle battery pack

Don't: /xray electric vehicle battery pack

If you type only the shortcut, ChatGPT will often explain what it thinks you mean instead of generating anything.

2. Upload a reference when you're changing an existing image

If you want the same subject, composition, product, character or structure, upload the image and say exactly what must stay unchanged.

Create image. Use the attached image as reference. Keep the exact subject, pose and composition, but convert the visual style to /clay.

If the reference is only for style, say that too:

Use the attached image only as a reference for illustration style, lighting and colour palette.

A reference can control style, composition, character, product, colour palette, lighting or material treatment. The more clearly you define what it controls, the better the result.

3. Name the icons and elements you want

The command tells ChatGPT how to visualise. Your keywords tell it what to include.

Instead of: Create image /iconset marketing

Use: Create image /iconset SEO, paid ads, email marketing, analytics, CRM, social media, content marketing, automation

This is the single easiest way to make image prompts more predictable.

4. Say exactly what should be labelled

Commands like /labelled, /xray, /anatomy, /blueprint and /exploded work far better when you name the components.

Create image /xray /labelled electric vehicle battery pack. Label battery cells, BMS, cooling plate, busbars, enclosure and high-voltage connector.

Avoid Create image /xray labelled battery — it leaves too much to guesswork.

5. Stack commands only when each one has a different job

Create image /xray /labelled /blueprint electric vehicle battery pack

Here /xray controls what is visible, /labelled adds annotations, and /blueprint sets the overall visual language. Three commands, three separate jobs.

Other combinations that work the same way:

Create image /miniature /isometric construction site

Create image /handwritten /sticknotes marketing strategy

Create image /clay /iconset AI tools

Create image /productshot /floatingproduct /rimlight perfume bottle

6. Always attach a subject

Never prompt with /miniature or /xray or /iconset alone.

Write: Create image /miniature construction site

Better still, add the key elements: Create image /miniature construction site, tower crane, excavator, workers, cement mixer, scaffolding, unfinished building

7. Use comma-separated keywords instead of long paragraphs

For most image styles, a clean keyword list is enough.

Create image /diorama futuristic AI office, robots, employees, holographic dashboards, meeting room, server racks, indoor plants

This works especially well with /iconset, /miniature, /diorama, /sticknotes, /stickerpack, /flatlay and /infographic.

8. Mention the background when it matters

Create image /3dicons CRM, analytics, email, calendar, automation, transparent background

Create image /productshot perfume bottle, clean white studio background

Create image /blueprint electric motor, dark navy blueprint background

This matters most when the image is going into a website, poster, carousel, deck or social post.

9. When editing, say what must not change

For a style transformation rather than a redesign, lock the original explicitly.

Keep the exact subject, pose, proportions, framing and composition. Do not add, remove or reposition any elements. Only change the visual style.

This one line dramatically reduces unwanted changes.

10. Add the format and composition

Useful instructions: 9:16 vertical, 1:1 square, 16:9 landscape, centered composition, transparent background, isolated object, top-down view, negative space on the left.

Create image /miniature construction site, 9:16 vertical, centered composition

8 mistakes that ruin image prompts

Don't use the shortcut alone. /iconset on its own gives ChatGPT nothing to work with. Use Create image /iconset finance, investment, savings, insurance, analytics.

Don't make ChatGPT guess the elements. Create image /iconset SaaS is a coin flip. Create image /iconset dashboard, integrations, API, billing, analytics, automation, security is a brief.

Don't reference an image you haven't uploaded. “Make it look like the previous image” is unreliable. If the exact reference matters, attach it and specify what to copy.

Don't mix conflicting styles. /watercolor /pixelart /photorealistic /clay /blueprint will fight each other. Pick commands that complement: /xray /labelled /blueprint, /miniature /isometric, /productshot /rimlight /reflection.

Don't use vague words when specific nouns exist. Replace “Create a cool AI image with technology” with Create image /miniature AI office, AI agents, CRM dashboard, analytics screens, workflow automation, email tools.

Don't assume ChatGPT knows what to copy from a reference. Instead of “Make something like this”, write: Use the attached reference only for the 3D clay material, pastel colour palette and soft lighting. Create a completely different subject.

Don't over-write simple prompts. A long paragraph is not automatically better. Instead of describing eight icons across 150 words: Create image /iconset SEO, paid ads, CRM, email, analytics, content, automation, social media. Consistent rounded 3D style. Transparent background.

Don't leave exact text ambiguous. When words need to appear in the image, put them in quotation marks: Create image /handwritten poster. Main headline: “Ideas Worth Testing”. Use only this exact text.

Copy-paste prompt formulas

Standard prompt

Create image + /command + subject + elements + visual direction + constraints

Create image /miniature construction site, tower crane, excavator, cement mixer, workers and scaffolding, realistic miniature photography, centered composition, 9:16 vertical

Create image /iconset CRM, analytics, email, calendar, database, automation, consistent clay style, transparent background

Create image /xray /labelled /blueprint electric vehicle battery pack, show battery cells, BMS, cooling system, busbars, enclosure and connectors

With a reference image

Create image + reference instruction + command + new subject/elements + what must stay consistent + what must not change

Create image. Use the attached image as the style reference. /iconset analytics, CRM, email, automation and calendar. Match the same 3D materials, lighting and colour palette. Keep all icons visually consistent. Transparent background.

For an image edit

Create image. Use the attached image. Keep the exact subject, pose, proportions, framing and composition. Do not add, remove or move anything. Only convert the image into /clay style.

Eight combinations worth saving

Technical visual: Create image /xray /labelled /blueprint electric vehicle battery pack

Educational visual: Create image /cutaway /anatomy human heart

Strategy visual: Create image /sticknotes /handwritten content marketing strategy, SEO, social, paid ads, email, CRM

SaaS icon pack: Create image /iconset /clayicons CRM, integrations, database, analytics, automation, API

Miniature scene: Create image /miniature /isometric futuristic AI office

Product visual: Create image /productshot /floatingproduct /rimlight luxury perfume bottle

Editorial explainer: Create image /annotation /circled landing page conversion audit

Product breakdown: Create image /exploded /labelled mechanical watch

The one rule to remember

Think about every prompt in four parts:

• Command = how it should look
• Keywords = what should appear
• Reference = what it should match
• “Create image” = tells ChatGPT to actually generate it

Once you start separating those four things, writing image prompts becomes much faster — and much more predictable.