# AI photoshoot from your photo: a series from one selfie

> An AI photoshoot is a series of shots a model draws from your selfie: the face stays yours, while the look, the place and the light come from a style or a prompt. Below: how to choose the source photo, 4 steps to the first series, which styles work better, why the face sometimes does not match, and whether you can try it for free.

Published 2026-09-30 · Updated 2026-09-30
URL: https://ai2face.com/blog/ii-fotosessiya

An AI photoshoot replaces the studio, the makeup artist and the photographer with one selfie. The model remembers the features of the face and draws you in any place, look and light: in a studio, in a café, on a beach or by a New Year tree. We made every example in this article from one photo taken at home, so you can see what holds the likeness and what breaks it. The frames were generated in [Nano Banana 2](/models/nano-banana-2) with no retouching.

## What an AI photoshoot from a photo is

It is the generation of new photographs of a person from their own shots. Current image models such as [Nano Banana](/models/nano-banana) take the photo as a reference and keep the likeness while changing clothes, pose, light and place. Google separately calls this a strength of the model: the same character in different outfits, angles and scenes without losing the likeness.

Earlier, this meant training a separate model on a dozen photos or more. Now one to three shots are enough, and each frame is one generation, with no training. The result depends on three things: the quality of the source photo, how precisely the scene is described, and the model.

![On the left, a selfie of a woman with a black bob on a sofa; on the right, four frames of the same woman: in a red dress on a burgundy background, in a café, on a beach at sunset, and by a New Year tree](https://ai2face.com/_astro/styles.C12l1qYS_Z1elJnb.jpg "One selfie from home and four styles: fashion, café, beach, a festive evening. Nano Banana 2")

## Which photo to upload

The likeness is about 80% decided by the source. The model does not see what is hidden by glasses, shadow or distance, and it fills that in by guesswork. We ran the same prompt on a good source and a bad one.

![Above, a good selfie and the studio portrait made from it, the same woman; below, a dark photo in glasses and a cap, and a portrait made from it that shows a different woman without bangs](https://ai2face.com/_astro/compare.BP6bvR8p_1KKNLI.jpg "The same prompt. From a sharp selfie, the same woman; from a dark frame in glasses, a different person. Nano Banana 2")

A checklist for a good source:

- **The face is toward the camera or in a slight turn,** and takes at least a third of the frame.
- **Daylight,** with no hard shadow on half the face and no sun behind the subject.
- **No glasses, cap or mask,** and hair does not cover the eyes.
- **No beauty filters:** the model will carry smoothed skin and enlarged eyes into the result.
- **One person in the frame:** on a group photo the model can mix features.

Two or three shots from different angles help when the series includes head turns and a profile.

## How to make an AI photoshoot from your own photo: 4 steps

In AI2Face the series is built in the Constructor — in layers, and a prompt is optional.

1. **Create a character.** Upload 1–3 selfies that pass the checklist above (up to 4 photos in total). The character stays in your library.
2. **Pick the scene.** Location, clothes and pose come from ready-made assets. You can upload your own photos of locations, clothes and poses on plans that allow it.
3. **Add a prompt if you need to.** Light, mood, time of day, lens — anything the assets do not cover.
4. **Pick the model and start.** The cost in credits is shown before you start. A frame that almost works can be edited: “change the dress to blue”, “make it evening”.

Layers work like a brief for a stylist: you assemble the frame from ready parts, and the model brings them into one photograph, with light and shadow.

![Four frames: the woman's selfie, an empty loft with a brick wall, an emerald silk dress on a hanger, and the result — the same woman in that dress by the loft window](https://ai2face.com/_astro/layers.BN7QNysM_Z1eB1J9.jpg "A frame from three layers: character, location and clothes. The loft, the chair and the dress are carried over exactly. Nano Banana 2")

## Which styles to pick

A style sets not only the place, but the light, and the light decides how much the frame looks like a real photograph.

| Style | What it is for | What to specify |
|---|---|---|
| Studio, business portrait | a résumé, a profile, a company site | a plain backdrop, a softbox, an 85 mm lens |
| Fashion, editorial | social feeds, a portfolio | a bright dress or a suit, hard light from above, a confident pose |
| Café, city | lifestyle, a blog | an action (“holds a cup”, “walks”), soft daylight |
| Beach, nature | vacation frames | golden hour, backlight, wind in the hair |
| A holiday: New Year, a birthday | cards, a holiday avatar | string lights, bokeh, warm evening light |
| Autumn | a seasonal series | the color of leaves and clothes, sun through the trees |

Ready text for each style, with the results, is collected in the article [photoshoot prompts](/blog/promty-dlya-fotosessii).

## Why the face does not match, and how to fix it

- **A bad source.** The most common cause; the example is above. Replace the photo.
- **No line about keeping the face.** If you write your own prompt, start with “keep the face, features and hair exactly as in the photo”.
- **The face is too small in the frame.** In a full-length shot the face has few pixels, and the likeness drops. For frames that matter, use a waist-up portrait or a close-up.
- **Heavy stylization.** Anime, 3D and oil paint change the proportions of the face — that is the style, not a failure.
- **The model refused.** Google's models do not create images of celebrities or recognizable brands. If a request is rejected, you can rerun it in the AI2Face generator on [Seedream](/models/seedream).

If a frame is almost right, do not generate it again: ask to fix one detail in the finished image. Editing techniques are in the guide [Nano Banana photo](/blog/nano-banana-foto).

## Which model to use for an AI photoshoot

| Model | Strength | When to use it |
|---|---|---|
| [Nano Banana 2](/models/nano-banana-2) | fast, up to 4K, holds the face across scenes | the main choice for a series |
| [Nano Banana Pro](/models/nano-banana-pro) | up to 5 people in the frame, complex light, text | couple and family frames, cards with lettering |
| Seedream | photorealism, a different set of limits | if a Google model rejects the request |

All three are in the AI2Face generator and switch in the same window. For a first series, take Nano Banana 2, and bring in Pro for frames with several people.

## Can you do an AI photoshoot for free

You can try it for free, but not finish a full series. Each generation costs the service money, so a “free” series usually means trial frames. In AI2Face, welcome credits are added after you sign up: enough to create a character and make the first shots in a few styles. Free access to Google's models directly is covered in the guide [Nano Banana free](/blog/nano-banana-besplatno).

## Is it safe to upload your photo

A photo of a face is personal data, so choose a service as carefully as you would a bank:

- upload only your own photos, or photos of people who gave consent;
- do not upload documents, photos of children, or frames you would not want in someone else's hands;
- check that the service does not publish your photos in a public feed.

In AI2Face, personal characters are tied to your account and are not visible to other users. Every image from Google's models carries an invisible SynthID mark — it shows the frame was made by AI.

## Conclusions

An AI photoshoot today is one good selfie and a few minutes instead of a day in a studio. The source decides it: a sharp photo in daylight, without glasses, keeps the likeness; a dark frame gives you a different person. A series is easier in the Constructor, with the character saved, than by uploading a photo into every request. Start with a studio or lifestyle frame on Nano Banana 2, then move to holiday and fashion scenes. If you need a real photographer for a large print or coverage of an event, a model will not replace them. For avatars, social feeds, a résumé and cards, it is already enough. You can try it in AI2Face on welcome credits.

## FAQ

### How do you make an AI photoshoot from your own photo?

Upload 1–3 sharp selfies, turn them into a character, pick a location, clothes and a pose, or write a prompt, and start. In AI2Face the character is saved, so the next frames do not need another upload.

### Can you do an AI photoshoot for free?

Partly. In AI2Face, welcome credits are added after you sign up — enough for the first frames. A series of dozens of shots is paid in any service: each frame costs money to generate.

### Can you do an AI portrait series without signing up?

In AI2Face, no: without an account there is nowhere to save the character or to add welcome credits. Services “without sign-up” usually give one or two frames and do not keep the face between generations.

### How many photos do you need?

One sharp photo, face toward the camera, is enough for the first frame. Two or three shots from different angles raise the likeness when the head turns. In AI2Face you can add up to 4 photos to a character.

### Why did the result look like a different person?

Most often the source: a dark frame, glasses, a cap, or a face far away. The model fills in what it cannot see. Replace it with a sharp selfie in daylight and add “keep the face as in the photo” to the prompt.

### Is it safe to upload your photo?

Upload only your own photos, or photos of people who agreed, with no documents and no other people's children. In AI2Face personal characters are tied to your account and are not visible to other users.

## Sources

- https://blog.google/products-and-platforms/products/gemini/nano-banana-tips/
- https://ai.google.dev/gemini-api/docs/interactions/image-generation
- https://support.google.com/gemini/answer/14286560