Bonus guide · Career tools
AI headshots, free first.
A new LinkedIn profile with an old selfie does not do the work. AI can produce a decent professional photo from casual pictures you already have on your phone. Start free. Pay only if you hit a wall.
What these services do
You upload 10 to 20 casual photos of yourself. The service fine-tunes an image model on your face, then generates new photos of you in settings you would never set up yourself: neutral studio backdrop, office, natural light. You pick the ones you like.
It is not a filter on top of one photo. It is a new image generated by a model that learned your face. That distinction matters when we talk about what can go wrong.
Free options to try first
Bing Image Creator (free with a Microsoft account)
bing.com/images/create. Free text-to-image. It will not recreate your face, but it's useful for placeholder professional-looking portraits or for practicing prompts. If you just need an avatar that looks like a person, this will do.
HuggingFace Spaces (free demos)
huggingface.co/spaces. A marketplace of free demos. Search for "headshot" or "portrait." Many demos let you upload a photo and generate variations. Quality varies a lot. Limits reset daily.
Google AI Studio (free)
aistudio.google.com. Google's free playground for their image models. Free with a Google account. Useful for generating illustrations, portraits, and backgrounds. Won't fine-tune on your face but works well for prompt-based portrait generation.
ChatGPT, Claude, or Gemini (free tiers)
All three of the major chatbots can generate images now on their free tiers (with limits). Ask for a "professional headshot of a person" with details and you'll get usable results. None of them will fine-tune on your face, but they're a solid free starting point.
Paid services that fine-tune on your face
Use these if the free options don't give you what you need. Prices change; check before you pay.
HeadshotPro
Polished experience. Good turnaround. Usually around $30 to $50 for a batch of photos. Tends toward clean corporate looks.
Aragon
More outfit and setting variation. Similar price range. Check the sample gallery for people who look like you before you pay.
Secta Labs
Another strong option. Pay attention to their privacy policy about how long they keep your training images.
Honest alternative: a real photographer in your city for a 30-minute headshot session often costs $50 to $150. Some libraries and community centers host free headshot events for job seekers. Ask the Urban League if any are coming up.
The workflow, start to finish
- Pick 15 to 20 casual photos of yourself. Mix of lighting, angles, and expressions. Not all selfies at arm's length. Include straight-on shots.
- Remove photos that confuse the model. Group photos, sunglasses, heavy filters, hats, photos from more than two years ago if your appearance has changed.
- Upload and wait. Most services take 30 to 90 minutes to fine-tune and generate a first batch.
- Review critically. Does this look like me, or does it look like a generic person with my face stapled on? If the second, try more varied training photos.
- Pick three to five you like. Resist picking your most flattering. Pick the ones that look most like you on a Tuesday morning.
- Download and use. Don't lie about it. If an employer asks "Is this an AI headshot?" the honest answer is yes. Most people won't care.
What to watch for in the output
- Hands that look wrong. Extra fingers, melted thumbs, watches reading backwards. Image models still struggle with hands.
- Company logos that never existed. The model will sometimes put a shirt or lanyard on you with a vague blurry logo. That logo isn't yours and shouldn't be in a professional profile photo.
- Text that is gibberish. ID badges, signs, book spines. The model does not know how to render real text. If there's a sign visible, assume it's wrong.
- Skin-tone accuracy. Many image models do better with some skin tones than others. If the output looks lighter or darker than you actually are, that's the model failing, not you. Regenerate or switch services. Don't accept a photo that misrepresents you.
- Jewelry and religious items. Cross necklaces, rings, hijabs, turbans. The model may drop or distort them. If it matters, include clear training photos that show it and check every output.
- Eyes. Some outputs have subtly off-center or unfocused eyes that read as "eerie." Spend the extra second checking.
- Age consistency. If you look 35 in training photos and the output looks 25, it won't match your real voice on a phone screen.
What NOT to keep in a profile AI just wrote for you
- Invented certifications. AI will happily say "CISSP-certified" if you ask for a polished profile. Do not keep it if it is not true. People verify in ten seconds.
- Inflated titles. "Director of Technology" when you were a sysadmin at a 20-person company. Hiring managers and former coworkers notice.
- Fake metrics. "Reduced costs by 47%." AI will cheerfully add numbers it has no way to verify.
- Awards that don't exist. "Featured in Forbes." Easy to search. Don't ship it.
- Language that isn't yours. If the AI rewrote your About section to sound like a venture capitalist and you're a medical assistant, the first phone screen will feel strange to the person on the other end. Keep your voice.
Privacy and the photos you upload
You are handing a service a detailed dataset of your face. Before you pay:
- Read the privacy policy. Look for how long they retain training data.
- Look for an explicit "do not use my images for further training" option.
- If the service is free, the photos are usually the product. Consider whether that trade is worth it for this use case.
If the service is for LinkedIn only and you're uncomfortable with retention: a real photographer in your city, or a community headshot event, is a fine alternative.