The mistake almost everyone makes is typing "professional headshot of me" and then blaming the tool for the waxy result. A real photographer makes about eight decisions before they press the shutter: the lens, the aperture, where the key light sits, what fills the shadows, how tight the crop is, what you are wearing, what is behind you, and how the file gets graded afterwards. A prompt that names all eight gets you a photograph. A prompt that names none of them gets you an avatar.
pick your routeThere are two genuinely different ways to do this, and picking the wrong one is why people give up. One trains a model on your face and hands you a hundred finished shots. The other writes one very precise photographic brief and renders it. They cost different amounts and they fail in different ways.
Upload a set of your selfies to a headshot service, it learns your face, and it returns dozens of finished shots across styles you picked in advance. Fastest path to a usable profile photo, paid, and you get variety without writing a single prompt.
Claude writes the photographic brief, a free image tool renders it, and you iterate one variable at a time. Slower, close to free, and you control every detail. This is the route worth learning if you want a repeatable look.
Aragon is the one I point people at when they want finished headshots rather than a hobby. You choose your outfits and backdrops first, upload a set of ordinary selfies, and it emails you when the batch is ready. Aragon's own site says six uploads is enough and to prioritise quality over quantity, and that you get back up to 100 headshots.
This is the step that decides everything. No prompt recovers from bad inputs, and five good reference photos beat fifty mediocre ones.
Claude writes the brief, but something else has to draw the picture. Two options that genuinely have a free entry point and accept your uploaded reference photos:
This is the single trick that makes the free route work, and almost nobody uses it. If you upload a selfie and ask for "a professional headshot", most image tools read that as permission to generate a new person who vaguely resembles you. If you upload the same selfie and ask it to edit that image, keeping the pixels of your face and changing only the light, the background and the clothes, it stays your face. Same tool, same photo, completely different result, because you changed what job you gave it.
Edit the attached photo of me. This is a retouch and relight job, not a new image: treat my face as fixed and unchangeable, and change only the environment around it.
Keep my exact facial structure, proportions, skin texture, freckles or marks, hairline and hair length precisely as they are in the source photo. Do not slim, smooth, whiten, age, de-age or beautify anything, and do not adjust my eye shape, nose or jaw. If you cannot make a change without altering my face, leave that change out and tell me instead of guessing.
What to change: relight the shot as if it were taken with soft daylight from a large window at 45 degrees to my left, with a gentle fill on the shadow side so nothing is crushed. Replace the background with a smooth neutral warm grey, softly out of focus. Replace what I am wearing with a well-fitted charcoal blazer over a plain white shirt, sitting naturally on my shoulders. Crop to a square head-and-shoulders frame with a little headroom above my hair. Finish with a neutral colour grade that looks like a real camera file rather than a filter, with visible pores and natural skin tone variation retained.
Paste this into any Claude, on any plan, at claude.ai. It makes Claude ask you the handful of things a photographer would ask before a shoot, then hands you a finished, paste-ready image prompt instead of a vague suggestion.
You are art-directing a professional headshot of me and your job is to write the image-generation prompt, not to give me advice. Before you write anything, ask me exactly five questions: what the headshot is for and who will see it, my industry and how formal it needs to read, what I actually own and would realistically wear, my colouring including hair, skin tone and eye colour, and whether I want it warm and approachable or cool and authoritative.
Once I answer, write me ONE image prompt of roughly 120 to 180 words as a single paragraph. It must explicitly name all of the following: the aspect ratio, the lens and aperture in photographic terms, the key light source with its direction and softness, the fill and any rim or hair light, how tight the crop is and where my eyeline goes, the wardrobe including fabric and colour, the background colour and how blurred it is, the expression, and the colour grade. Include a hard instruction to keep my face, bone structure, skin texture, freckles or marks, and hairline exactly as they appear in my reference photos, and to avoid smoothing, slimming or beautifying anything.
End the prompt with a short negative list of what to avoid: plastic or waxy skin, over-sharpened eyes, symmetrical fake teeth, floating hair, warped ears, and the glossy over-lit look of stock corporate photography. Then, separately from the prompt, give me three one-line variations I can run afterwards to change only the background, only the wardrobe, and only the mood.
These are the levers Claude is pulling in that brief. Learn them once and you can write your own prompts for anything with a face in it.
Four complete looks. Upload your five to ten reference photos in the same message, then paste one of these. Each one already names all six dials.
1. The LinkedIn default. Clean, neutral, works in any industry, and the one to make first.
Using the attached photos of me as the reference for my face, create one square 1:1 ultra-realistic professional headshot of the same person. Shot on an 85mm lens at f2, head and shoulders, a small amount of headroom above the hair, eyes tack sharp and looking directly down the lens. Key light is a large softbox positioned 45 degrees to camera left and slightly above eye level, soft and wrapping, with a gentle fill on the shadow side and a subtle rim light behind my right shoulder to separate me from the background. Wardrobe: a well-fitted charcoal blazer over a plain white cotton shirt, no pattern. Background: smooth neutral warm grey, softly out of focus. Expression: relaxed, open, a small genuine smile that reaches the eyes.
Keep my face, bone structure, skin texture, freckles or marks, and hairline exactly as they appear in the reference photos. Do not smooth, slim, whiten or beautify anything. Retain visible pores and natural skin tone variation. Finish with a neutral colour grade like a real digital camera file, not a filter.
Avoid: plastic or waxy skin, over-sharpened eyes, unnaturally symmetrical or over-white teeth, floating or fused hair strands, warped ears or glasses frames, and the glossy over-lit look of stock corporate photography.
2. The warm founder shot. For an about page, a personal brand, or anywhere "approachable" matters more than "corporate".
Using the attached photos of me as the reference for my face, create one 4:5 ultra-realistic portrait of the same person for a personal website about page. Shot on an 85mm lens at f1.8, from just below eye level so it feels grounded and confident, framed from mid-chest up with my body angled slightly away and my face turned back toward camera. Key light is warm, low-angle late-afternoon window light from camera right, soft with a visible but gentle falloff across the far cheek, plus a soft bounce fill. Wardrobe: a soft oatmeal knit or a relaxed linen shirt, sleeves pushed up, nothing corporate. Background: a warm, lived-in interior in cream and wood tones, heavily out of focus, with one soft highlight bloom in the top corner. Expression: mid-conversation, a real unforced smile, as if I have just finished saying something.
Keep my face, bone structure, skin texture, freckles or marks, and hairline exactly as they appear in the reference photos. Do not smooth, slim or beautify. Retain real skin texture and any asymmetry, because that asymmetry is what makes it read as a photograph.
Avoid: plastic skin, a stiff posed stance, over-saturated orange grading, floating hair, and any hint of stock-photo styling.
3. The editorial film look. For press features, speaker bios, podcast artwork, or a profile photo with some personality.
Using the attached photos of me as the reference for my face, create one 4:5 ultra-realistic editorial portrait of the same person in the style of a magazine feature. Shot on a 50mm lens at f2 on 35mm film, with fine natural grain and slightly muted highlight rolloff. Framed from the chest up, my head turned three quarters to camera left with my eyes coming back to the lens. Key light is a single hard-ish source from high camera left creating a defined shadow along the jaw, with almost no fill so the shadow side stays genuinely dark, and a faint practical light glow somewhere behind me. Wardrobe: a plain black or deep navy high-neck top, no visible logos or jewellery. Background: a deep, unlit interior falling off to near black. Expression: still, unsmiling, direct and calm.
Keep my face, bone structure, skin texture, freckles or marks, and hairline exactly as they appear in the reference photos. Do not smooth, slim or beautify. Preserve real skin texture including any lines and pores, since this look depends on texture being visible.
Avoid: plastic skin, digital over-sharpening, HDR-style shadow lifting, heavy vignetting, fake film burn or light-leak overlays, and unnaturally symmetrical features.
4. The environmental work shot. A wider frame that shows what you actually do, which is the one most people are missing.
Using the attached photos of me as the reference for my face, create one 16:9 ultra-realistic environmental portrait of the same person at work. Shot on a 35mm lens at f2.8, wide enough to show a real workspace around me, framed from the waist up with me positioned slightly off centre and the room readable behind. Key light is soft daylight from a large window out of frame to camera left, with the room's own practical lights adding warm accents in the background. Wardrobe: smart casual, a structured neutral shirt or a light knit, the kind of thing someone actually wears to work. Background: a genuine desk environment with depth, gently out of focus but still legible. Expression: mid-task and natural, looking up toward camera as if I have just been interrupted, a slight smile.
Keep my face, bone structure, skin texture, freckles or marks, and hairline exactly as they appear in the reference photos. Do not smooth, slim or beautify. Keep my hands, if visible, anatomically correct with five fingers each and no distortion.
Avoid: plastic skin, an empty sterile stock office, floating or duplicated objects on the desk, unreadable text on screens rendered as gibberish (leave screens dark or blurred instead), and the over-lit look of corporate stock photography.
Once you have a shot that is close, do not rewrite the whole prompt. Change one variable per message in the same chat, so you can see what each change actually did.
Zoom to 100 percent and check these five before the image goes anywhere near your profile. This is where AI headshots get caught.
Claude, free to start
Reve and its pricing page
Aragon, its pricing, the free generator and its security policy