A reliable AI headshot prompt starts from your own photo, locks your face first, then sets wardrobe, background, light, framing, expression and shape, with a short avoid list. Below are 20 fill-in formulas by role, and a table showing how to write the avoid list in ChatGPT, Gemini, Midjourney and Flux. They are templates to adapt, not fixed recipes.
A good AI headshot prompt locks your face first, then describes wardrobe, background, light, framing, expression and shape, in that order: "Keep my face, hair and features unchanged. Head-and-shoulders headshot, navy blazer, plain light grey backdrop, soft window light from the left, relaxed smile, square." Add a short avoid list for the failures you actually see, and write it in the dialect your tool understands.
This pack gives you 20 fill-in formulas by role, from corporate suits to clinicians and founders, each with an avoid list, plus notes on how to adapt them for ChatGPT Images, Gemini, Midjourney and Flux. They are templates to adapt rather than recipes with fixed results: output varies by tool, model version and the photo you start from. For choosing a tool or a paid headshot service in the first place, see our complete guide to AI headshots.
The AI headshot prompt formula
Every formula below follows the same seven slots. Knowing the slots matters more than memorising any single prompt, because it lets you fix one part without breaking the rest.
- 1Lock
Keep my face, hair and features unchanged. Always first.
- 2Framing
Head and shoulders for profiles; waist-up for bios and websites.
- 3Wardrobe
One specific outfit in plain colours, no logos.
- 4Background
A colour, or one blurred scene type. Simple survives the crop.
- 5Light
Soft, from one named side. Window light is the safe default.
- 6Expression
One phrase: relaxed smile, composed, warm.
- 7Shape and avoid list
Square or 4:5, then the few failures to exclude.
Two rules hold across tools. First, start from a recent, unfiltered photo of yourself taken at eye level in even light; a text-only prompt produces a stranger, however detailed it is. Second, change one slot per round. If the background is right and the jacket is wrong, ask for a different jacket only.
The 20 formulas
Replace the bracketed parts with your own choices. The third column is the avoid list: write it as instructions in ChatGPT and Gemini, after --no in Midjourney, and as positive descriptions in Flux (see the dialect table below).
| Role | Formula (edit the brackets) | Avoid list |
|---|---|---|
| 1. Corporate suit | Keep my face, hair and features unchanged. Head-and-shoulders corporate headshot, [charcoal suit, white shirt, no tie], plain [light grey] backdrop, soft key light from the [left], gentle fill, calm confident expression, square. | tie pattern, shiny fabric, harsh shadows, airbrushed skin |
| 2. Business casual | Keep my face unchanged. Head and shoulders, [navy knit polo], softly blurred modern office behind me, natural window light, relaxed smile, square. | logos, laptop screens, busy desks, glow |
| 3. Startup founder | Keep my face unchanged. Waist-up portrait, [plain dark t-shirt and overshirt], blurred bright workspace, daylight from a large window, direct look at camera, half smile, [4:5]. | hoodie logo, neon, wide-angle distortion |
| 4. Creative director | Keep my face unchanged. Head and shoulders, [black turtleneck], solid [terracotta] backdrop, single soft light from the side, thoughtful expression, square. | heavy grain, colour cast on skin, props |
| 5. Physician | Keep my face unchanged. Head and shoulders, [white coat over a light blue shirt], plain [soft white] background, even soft light, warm reassuring smile, square. | stethoscope text, hospital signage, invented badges |
| 6. Nurse or clinician | Keep my face unchanged. Head and shoulders, [navy scrubs], softly blurred clinic corridor, even daylight, friendly smile, square. | name badges, logos, medical equipment in focus |
| 7. Dentist or therapist | Keep my face unchanged. Head and shoulders, [soft blue blouse or shirt], calm neutral background, warm soft light, gentle open smile, square. | over-white teeth, dramatic shadows |
| 8. Lawyer | Keep my face unchanged. Head and shoulders, [dark suit], blurred bookshelf of plain spines, soft directional light, composed expression, square. | readable book titles, gavel, flags |
| 9. Finance or advisor | Keep my face unchanged. Head and shoulders, [navy suit, light shirt], plain [mid grey] backdrop, classic studio light, steady eye contact, square. | charts, money, gold accessories |
| 10. Real estate agent | Keep my face unchanged. Waist-up portrait, [tailored blazer], softly blurred bright home interior, natural light, warm approachable smile, [4:5]. | sold signs, house numbers, fake keys |
| 11. Engineer or developer | Keep my face unchanged. Head and shoulders, [plain crew-neck sweater], blurred studio wall in [cool grey], soft light, relaxed expression, square. | code on screens, glowing monitors, headset |
| 12. Academic or researcher | Keep my face unchanged. Head and shoulders, [tweed or plain jacket], blurred library or lab background, soft window light, thoughtful half smile, square. | readable text, lab hazard symbols |
| 13. Teacher or trainer | Keep my face unchanged. Head and shoulders, [cardigan or plain shirt], softly blurred bright classroom, daylight, open friendly smile, square. | children in frame, readable boards |
| 14. Consultant outdoors | Keep my face unchanged. Head and shoulders, [light blazer], blurred city street at golden hour, warm side light, confident smile, square. | other people, traffic, lens flare across face |
| 15. Speaker bio | Keep my face unchanged. Waist-up, [dark blazer], dark backdrop with a soft rim light, engaged expression as if mid-conversation, [4:5]. | microphone, stage text, audience |
| 16. Author photo | Keep my face unchanged. Head and shoulders, [knit sweater], softly blurred window seat, natural light, quiet confident look, [4:5]. | book covers, readable titles |
| 17. Sales or client-facing | Keep my face unchanged. Head and shoulders, [smart casual shirt], bright plain [pale blue] backdrop, even soft light, wide genuine smile, square. | teeth whitening, glossy skin |
| 18. Executive black and white | Keep my face unchanged. Head and shoulders, [dark suit or turtleneck], black and white, low-key side light, dark background, composed expression, square. | crushed shadows on face, heavy contrast |
| 19. Fitness or wellness coach | Keep my face unchanged. Waist-up, [plain fitted athletic top], blurred bright studio, soft daylight, energetic smile, [4:5]. | body reshaping, muscle enhancement, logos |
| 20. Remote worker | Keep my face unchanged. Head and shoulders, [plain sweater], softly blurred tidy home office with plants, window light, relaxed smile, square. | cluttered shelves, screens, ring light reflections |
The avoid lists target the problems that show up most in professional portraits: invented text and badges, logos, beauty retouching that changes the face, and backgrounds that compete with it. Add to them only when you see a specific failure; a long pasted list tends to cause more problems than it fixes. Our guide to negative prompts for realistic people explains which terms are worth keeping.
Weak versus strong headshot prompts
- Professional headshot of a man, 35, handsome, perfect skin, 8k, ultra realistic
- No reference photo
- No light direction or background
- Asks for flawless, which reshapes faces
- Keep my face, hair and features unchanged. Head and shoulders, charcoal blazer, plain light grey backdrop, soft window light from the left, relaxed smile, square
- Starts from your own photo
- One slot per instruction
- Short avoid list: airbrushed skin, logos
The weak version describes a generic person and invites the tool to invent one. The strong version describes a photo session: who (you, unchanged), what you wear, where you stand, how the light falls. That is the difference between a stock-photo face and one your colleagues recognise.
Headshot prompt for Midjourney
Midjourney reads prompts as descriptions, with parameters at the end. For a headshot of yourself:
professional head-and-shoulders portrait of the person in the reference image, charcoal blazer, plain light grey backdrop, soft window light from the left, relaxed natural smile --ar 1:1 --no logo, text, watermark, illustration
- Your face: in Versions 8.1 and 8.2, attach your photo to the Edit model, which takes up to four reference images and also accepts written instructions; it replaces Omni Reference and Character Reference. In Version 7, use Omni Reference with one image. Midjourney notes that fine details such as freckles may not match exactly (checked October 2026).
- Shape: Midjourney images start square;
--ar 4:5gives a portrait crop. The aspect ratio docs list 1:1 as popular for social media profile pictures, and the Edit model docs say it matches the first uploaded image's ratio unless you add--ar. - Exclusions:
--notakes a comma-separated list. Midjourney's moderation reads each word separately, so avoid phrases such as "modern clothing" after--no; describe the clothing you want instead (checked October 2026).
Midjourney's rules for uploaded images say you must have the rights to them and must not manipulate images of people in abusive ways. Use your own photo, or a photo of someone who agreed. More style vocabulary lives in our Midjourney prompts guide.
Professional photo prompt AI dialects: one formula, five tools
| Tool | How to add your face | How to write the avoid list | Shape |
|---|---|---|---|
| ChatGPT Images | Upload your photo, paste the formula as an instruction | Write the avoid list as plain instructions: "Do not change my face shape; do not add jewellery" | Pick square or 4:5 in the aspect ratio picker |
| Gemini (Nano Banana 2) | Upload your photo, paste the formula as an instruction | Plain instructions, as above | State the shape in the prompt |
| Midjourney V8.1 / V8.2 | Attach your photo to the Edit model (up to four references); instructions allowed | --no item, item at the end | --ar 1:1 or --ar 4:5 at the end |
| Midjourney V7 | Omni Reference with one image (--oref plus URL in Discord) | --no item, item at the end | --ar 1:1 or --ar 4:5 |
| FLUX models | Your reference workflow (for example a Kontext edit) | Not supported: rewrite each item as a positive description | Set the size in the tool or API |
Sources, checked October 2026: OpenAI Images in ChatGPT, Google Generate images with Gemini Apps, Midjourney docs linked above, Black Forest Labs guide to working without negative prompts.
In Flux, the avoid list has to become description. Black Forest Labs says FLUX models do not support negative prompts and that negation can backfire: writing "without glasses" can make glasses more likely. So "airbrushed skin" becomes "natural skin texture with visible pores", and "logos" becomes "plain unbranded clothing". Our Flux Kontext guide covers editing from a reference photo in Flux.
LinkedIn photo prompt
For LinkedIn, use formula 1, 2 or 11 and tighten the framing. LinkedIn's own photo tips suggest your face should fill about 60% of the frame, from the top of your shoulders to just above your head, and its profile photo guidelines say the photo must reflect your likeness (checked October 2026). A LinkedIn-ready version:
Keep my face, hair and features unchanged. Close head-and-shoulders headshot, face filling most of the frame, [navy blazer over a white top], plain [light grey] background, soft window light from the [left], relaxed natural smile, square. Do not change my face shape, age or skin tone. Do not add jewellery, logos or text.
Then export a square file within LinkedIn's upload limits. Our AI profile picture guide lists the official size and file rules for LinkedIn and six other platforms.
How to adapt a formula to your role
None of the 20 will fit you exactly, and that is the point of the brackets. Three adaptations cover most cases:
- Industry dress codes. Swap the wardrobe slot for what people in your field actually wear on a good day, not a costume of the profession. A designer in a suit reads as wrong as a banker in a hoodie.
- Your own clothes and accessories. If you always wear glasses, a headscarf or a particular style, name it in the wardrobe slot so the tool keeps it.
- Brand colour. Use the background slot for a brand colour, but keep it muted so it does not tint skin.
If you want to go further than one headshot, into a consistent AI character across hundreds of images, our AI Influencers program covers reference sets, style systems and disclosure in depth.
Before you publish
- Started from your own recent photo, or one from someone who agreed
- Face shape, age, skin tone and features unchanged
- No invented text, badges, logos or jewellery
- Background simple enough to survive a circle crop
- Light comes from one side and looks natural on the skin
- Square or 4:5, exported within the platform limits
- Compared side by side with a real photo of you
AI headshot prompts: FAQ
What is a good AI headshot prompt?
A good AI headshot prompt starts from your own uploaded photo and says what must stay fixed before what should change: keep my face and features unchanged, then wardrobe, background, light direction, framing and expression, then the shape, usually square. Add a short avoid list for the failures you see, such as plastic skin or an extra accessory. Treat any prompt as a template and expect two or three rounds of small edits.
How do I write a headshot prompt for Midjourney?
Describe the portrait in plain words, then put parameters at the end: --ar 1:1 for a square crop and --no followed by things to exclude. To use your own face, attach your photo to the Edit model in Versions 8.1 and 8.2, which takes up to four reference images, or use Omni Reference in Version 7. Midjourney warns that fine details may not match the reference exactly. Checked October 2026.
Does Flux support negative prompts for headshots?
No. Black Forest Labs says FLUX models do not support negative prompts and advises describing what you want instead. For a headshot, turn each exclusion into a positive phrase: natural skin texture with visible pores instead of no plastic skin, plain light grey background instead of no clutter, and so on. Writing without glasses can even make glasses more likely. Checked October 2026.
What is the best prompt for a LinkedIn photo?
Use the clean professional formula: keep my face unchanged, head and top of shoulders, face filling most of the frame, plain light grey or soft blurred office background, soft window light, a navy or charcoal top, relaxed natural smile, square. LinkedIn's own tips suggest your face should take up about 60% of the frame, and its guidelines require the photo to reflect your likeness. Checked October 2026.
Can I use these prompts in ChatGPT or Gemini?
Yes. Upload your photo, paste the formula and fill the brackets. Neither tool documents a separate negative-prompt field in its help pages, so write the avoid list as plain instructions, such as do not change my face shape or add jewellery. OpenAI says ChatGPT Images has an aspect ratio picker, and Google lists uploading images to edit among Nano Banana 2 features in Gemini. Checked October 2026.
Why does my AI headshot not look like me?
Usually because the prompt let the tool change your face. Put keep my face, hair and features unchanged first, use a recent unfiltered photo taken at eye level in even light, and remove words such as flawless, perfect skin or model, which push towards a generic face. Change one thing per round. If a result alters face shape, age or skin tone, discard it rather than editing it further.
Prompts are the start. Consistent AI faces are the system.
AI Influencers, included in All Access, covers reference sets, consistent characters, image and video workflows and disclosure, with the other three programs, live coaching and the private community in one subscription.
Get new prompt formulas free
Join the free Telegram channel for the prompts and tools we are using right now.