What an AI headshot actually is
A headshot is a deliberate, narrow shot of a person — shoulders up, face clearly readable, framing optimized for cropping into circles, squares, and small tiles. It is not a portrait of an environment or a moment; it is a portrait of identity. For decades, the path to a usable headshot was: book a photographer, sit through twenty minutes of light tests, accept whatever frame they shipped, pay between two hundred and twelve hundred dollars, and reuse the file for two or three years. AI generation in 2026 short-circuits the path — you send selfies, the model anchors on your face, and a few seconds of inference produces a studio-grade frame.
Three things have changed since 2024 that make this practical instead of novelty. Identity preservation in image-conditioned diffusion models is now stable across reference inputs of varying quality: the face you bring is the face you get back, with consistent geometry across runs. Light and framing controls are now expressible as templates rather than long prompts: a Smart Template names a target style, the system handles the underlying lighting and composition prompt fragments. And per-image cost has dropped under three cents wholesale on the leading models, so a per-portrait price under a dollar can include retries, refunds, and the cost of running the bot.
@vustPortraitBot ships three Smart Templates — Founder Studio, Lifestyle Founder, Boardroom Formal — each tuned for a different audience surface. The page you are reading exists because most people Googling AI headshot generator want to see results before paying, and want to know what the contract is. The page is not a generation surface; the bot is.
Identity preservation, in practice
Identity preservation is the bar where AI portrait products either succeed or earn the "looks like someone else" complaint. The model anchors on facial geometry — interocular distance, nose bridge, jaw shape, brow ridge, lip line. Skin texture, hair, and expression are partially anchored and partially synthesized; this is why the same selfie produces a slightly different result on each run. What stays constant is the recognizable identity. What varies is the moment.
You can tilt the result toward your favor by sending the right reference photos:
- Three or four selfies from the same lighting day. Mixing morning kitchen light with evening street light creates inconsistent skin tone signals and the model averages them — sometimes flatteringly, sometimes not. Pick one lighting condition, take a few selfies in it, and stop.
- Eyes open, looking at the camera. The model uses the iris to anchor gaze; closed-eye or sideways selfies produce headshots that look slightly off without you being able to name why.
- Neutral or gentle expression. A toothy laugh in the reference biases the model toward an expression that may not match the studio-lit template; a soft expression keeps the model focused on identity rather than mood.
- Frame the face, not the room. Close enough that the face fills most of the vertical, with some neck and shoulder visible. Selfie-arm-extended works fine; mirror selfies work if the lighting is good and the camera is tilted slightly above eye level.
We do not require studio reference photos. The system is built around what people actually have on their phones — a few selfies in a coffee shop, a few selfies in a hotel room. The Smart Templates cover the rest.
What each Smart Template owns
Founder Studio is the workhorse. Neutral grey backdrop, soft key light from the upper left, slight rim light to separate hair from background, a touch of warmth in the skin tone. Crop is shoulders-up, head occupies roughly the upper half of the frame, eye line is on the upper third. This is what you reach for when the result will be cropped into a LinkedIn circle, dropped into an About-page team grid, used as a podcast guest tile, or sent to a journalist asking for a headshot. It reads as professional without reading as corporate.
Lifestyle Founder is warmer. The backdrop is suggested rather than literal — usually a softer blur with hints of warm color. Light comes from a window-side angle, color grade is a few degrees warmer than Founder Studio, and the angle of the head suggests a brief moment of conversation rather than a fixed studio pose. This template fits product, design, content, and creative roles where the audience expects approachability. It also works well when the photo will sit next to long-form writing — a Substack avatar, a personal website, a guest essay byline.
Boardroom Formal is the strict one. Suit, tie or formal collared option, controlled symmetric light, traditional headshot crop. Backdrop is darker than Founder Studio, framing is tighter, the eye line sits on the lower third — the composition style of executive photography. Reach for it when the audience is finance, law, traditional consulting, board work, or any context where formality is the unspoken expectation. It reads as serious without reading as cold, but the trade-off is less personality than the other two.
When AI is enough, and when it is not
For 2026 the honest answer is: AI is enough for almost every professional surface a non-celebrity adult uses. Speaker bios, podcast guest tiles, conference badges, About sections, team grids, press kits, podcast cover thumbnails, social profiles — all surfaces where the portrait is consumed at less than 600 pixels in the longest dimension. At those sizes, the difference between an AI portrait and a paid photographer's frame is usually invisible to the audience.
AI is not enough yet for two narrow cases. The first is a print campaign or magazine editorial where the image will be reproduced at high resolution and reviewed shot-for-shot by an art director. The second is any context where the photographer is a deliberate creative voice — fashion editorial, conceptual portraiture, brand campaigns — and the choice of image is part of the story. For both of those, hire a photographer.
The middle case is the interesting one. Suppose you run a modest company and need a portrait for the company About page. A photographer charges three to six hundred dollars; the portrait will be excellent and last three years. An AI portrait costs five crystals and renders in under a minute; the portrait will be very good and you can re-render any time wardrobe, hair, or context changes. The break-even is whether you re-shoot more than once every three years. Most people do.
Common failure modes and how to avoid them
A handful of recurring problems trip up first-time AI portrait users. Naming them upfront avoids most of them.
Identity drift on tight crops. Boardroom Formal is the tightest template; with only one reference selfie, the model can drift on jaw or hairline. Send three or four selfies for tight templates; one is fine for Founder Studio.
Wardrobe mismatch. The templates assume business-leaning dress. If you send a selfie in a tank top and pick Boardroom Formal, the model will dress you in a suit; this usually looks fine, but the neckline boundary occasionally renders as visibly synthesized. To avoid this, send selfies wearing something close to what the template expects — a collared top for Founder Studio and Boardroom Formal, a clean t-shirt or sweater for Lifestyle Founder.
Glasses. Single-frame glasses render reliably. Reflective lenses or strong frame textures sometimes confuse the model. If your daily glasses are minimalist, you are fine. If they are large or heavily decorated, send one selfie with glasses and one without; the model will choose the cleaner option.
Hair-color drift. The model occasionally lightens or darkens hair by a single tone. If exact hair match matters — a known brand color you've been using on every public surface — render two takes and pick the closer one. Pro portrait credits cover this while credits remain.
Background expectations. Founder Studio and Boardroom Formal use neutral backgrounds; Lifestyle Founder hints at a warmer environment. The MVP does not let you pick a specific environment such as "your office" or "in front of a logo." Background swap is on the backlog, not in MVP.
Format specs and how to use them
Output is a JPEG at the native resolution of the model. For 2026 that is large enough to use as a high-resolution profile picture and most company-page modules. The aspect is 4:5 portrait by default. If you need a square version for LinkedIn, save the file and use any image tool — Mac Preview, Photoshop, Canva, even Telegram itself — to crop into a square. Center-of-face cropping is built into the composition, so most square crops work without losing the eyes.
For LinkedIn specifically, upload the square crop directly. LinkedIn's profile picture render is 400×400 minimum and accepts 8 MB; our portraits are well under both limits. The platform crops into a circle on display; our composition leaves margin around the head so the circle does not clip an ear or chin.
For company About pages, upload the original 4:5. Most CMS modules either accept 4:5 directly or center-crop to fit a tile. If your team page uses a fixed crop ratio, render once and use any cropping tool to match.
For press kits, send the full-resolution JPEG. Some publications request 1500×1500 or larger; the output comfortably exceeds typical print-web requirements.
What we explicitly do not claim
The page you are reading does not promise anything about identity verification, employment outcomes, recruiter response rates, or beauty improvement. AI portraits are images; they cannot validate who you are. They cannot get you hired. They cannot make you objectively more attractive. What they can do is render a frame that is consistent with how you want to be seen on professional surfaces, in a few seconds, for a few crystals.
The bot also adds an explicit AI-generated disclosure caption to every delivered portrait. Use the picture where AI imagery is allowed and add a one-line disclosure on the surface itself when the platform expects it. Treat the caption as a starting point, not as a substitute for the disclosure norms of the platform you upload to.
Privacy is enforced in code: image bytes live only in memory during generation, and only Telegram file references and bounded metadata are persisted. The contract is documented in the security baseline doc and the code reflects it.
Practical workflow
A clean run looks like this. Open @vustPortraitBot in Telegram. Send three selfies — same lighting day, eyes open, gentle expression. Tap the template that fits the audience surface you are filling. Confirm. Within thirty seconds the portrait lands in the chat with an AI-generated disclosure caption. Save the file, upload it to the surface you intended, and write a one-line disclosure in the bio if the platform requires it.
Five crystals leaves your wallet for the delivered portrait. If generation or delivery fails after the debit, the wallet is refunded automatically and the bot reports the refund. If you want to A/B test wardrobe or template, switch to Pro — the 30-day Pro pass includes 12 portrait credits, and longer passes include more. Beyond that the workflow is the same.
That is the whole product. The web page you are reading exists so the contract is visible and the result is searchable. The work happens in Telegram.