# Multiple Characters in One Image: How to Use It in AI2Face

> Since October 21, 2026, AI2Face lets you put multiple characters in one image: attach them to your prompt and call them by name — “@Sam teaches @Emma and @Leo.” It works for photos in every model and for video wherever the model accepts reference images. Below is every step with screenshots: creating a character, the Characters tab, @ tags and your own background via @1.

Published 2026-10-21 · Updated 2026-10-21
URL: https://ai2face.com/blog/neskolko-personazhej-v-odnom-kadre

Until now, an AI2Face image could include only one character — the one selected in the Builder. Now you can attach as many characters to a prompt as the model's limit allows and refer to them by name in the text. This solves the main problem with group shots: the model no longer mixes up whose face goes where. Below is a step-by-step walkthrough with interface screenshots. We tested how the feature performs across models in our guide on [how to combine photos with AI](/blog/obedinit-foto-nejroset).

## What's new: multiple characters in one image and @ tags

Since October 21, 2026, the AI2Face generator has:

- **a Characters tab in the “+” panel** — attach several characters to a prompt at once, just like regular photos;
- **@ tags in prompts** — characters by name (@Emma), photos by number (@1, @2), with suggestions when you type “@” and highlighting in the field;
- **a server-side master prompt** — AI2Face's proprietary algorithm turns your tagged text into detailed instructions for the model: who is in which photo and what to preserve;
- **every image model** — [Nano Banana](/models/nano-banana), [GPT Image](/models/gpt-image), [Seedream](/models/seedream) and the rest understand tags the same way;
- **every video model with references** — the same characters and tags in your clips.

## How to create a character

A character is a name plus 1–4 photos of one person. Open Characters in the top menu and click Create a character. In the New character window, enter a name — it also becomes the tag you use in prompts.

![The New character window with a name field over a list of characters Sam, Leo and Emma](https://ai2face.com/_astro/screen-new-character.Cyr8h4ke_1lR7b4.jpg "Characters → Create a character: the name becomes the @Name tag")

After you create it, the character page opens. Click Add photo and upload 2–4 shots. Good choices: a front-facing face, a three-quarter view and a full-length shot in daylight. Avoid group photos, covered faces, filters and masks, and dark or blurry shots — AI2Face shows the same tips under the photos.

![Emma's character page with two photos — a window-lit portrait and a shot in a park — and an Add photo tile](https://ai2face.com/_astro/screen-character-photos.f-dY5M59_1h2LHn.jpg "Two photos from different angles: AI2Face builds the character profile for the model from them")

With two or more photos, AI2Face merges them into a single character profile: the model sees the person from several sides, and the character counts as one slot toward the photo limit.

## How to attach characters to a prompt

In the generator, click “+” to the left of the prompt and open the Characters tab. Click a tile to attach a character; click again to remove it. The panel stays open, so it's easy to select several in a row.

![The “+” panel on the Characters tab: Sam, Leo and Emma are selected, with their thumbnails above the prompt below](https://ai2face.com/_astro/screen-characters-tab.3Ve2QpyU_2piboS.jpg "The Characters tab: three selected characters appear above the prompt right away")

Characters sit above the prompt alongside your other photos. The counter on the right — for example, 3/14 — shows how many photos will go to the model and how many it accepts. You can drag thumbnails to reorder them, and the cross on a thumbnail removes the character from the prompt.

## How to name a character in the prompt: @ tags

Type “@” at the start of a word to open a list of everything attached to the prompt: characters and photos. Start typing a name to narrow the list, then press Enter or Tab to insert the tag.

![A dropdown with @Sam, @Emma and @Leo above the prompt reading “@Sam teaches @”](https://ai2face.com/_astro/screen-mention-list.I1x3MPoW_Z2liKGI.jpg "Typing “@” opens a list of characters and photos")

Tags are highlighted in the field, and hovering over one shows its photo — an easy way to check that “@Emma” is the right Emma.

![The prompt “@Sam teaches @Emma and @Leo to make fresh pasta…” with three highlighted tags](https://ai2face.com/_astro/screen-prompt-mentions.DPTaSZOb_Z19fuwL.jpg "The finished prompt: every person is named with a tag")

![An older man teaches a red-haired woman and a bearded man to make pasta in a rustic kitchen](https://ai2face.com/_astro/kitchen.BK3losud_UV4d0.jpg "The result of this prompt in Nano Banana Pro, 3:2, 2K")

Tag rules:

- **Characters by name,** photos by their number among photos: @1, @2. The number is shown on the thumbnail.
- **Not case-sensitive:** @emma and @Emma are the same tag.
- **Spaces in a name** become “_”: a character named “Anna Maria” is @Anna_Maria.
- **Duplicate names** get a suffix: a second Emma becomes @Emma_2.
- **Removed a character from “+”** — its tag in the text turns into the plain name without “@”. Removed a photo — its tag disappears and the numbers of the following photos shift.

What happens when you click Generate: AI2Face's proprietary server-side algorithm parses the prompt, matches each tag to its photo and assembles a master prompt — detailed instructions for the model on who is who and what to preserve in each face. You don't need to describe people by their clothes or explain which photo belongs to whom: a short “@Sam teaches @Emma and @Leo” is enough. This is AI2Face technology, and it works the same way with every model in the generator, from Nano Banana to GPT Image. The image card keeps your tagged text.

## Your own background, pet or object via @1

You can add any photo next to your characters: an interior, a landscape, a pet, an object. Attach it with “+” from Uploaded or Generated and refer to it by number. In the screenshot below, a photo of a terrace became @1, and the prompt asks to seat all three people at the table.

![Characters Emma, Leo and Sam plus a terrace photo tagged @1 above the prompt, which reads “on the terrace from @1”](https://ai2face.com/_astro/screen-photo-ref.DosVLz40_Z23iqcw.jpg "The terrace photo is attached fourth and gets the @1 tag")

![On the left, an empty mountain-cabin terrace with breakfast; on the right, the same table and view with Sam, Emma and Leo seated](https://ai2face.com/_astro/terrace.Dvhn58yq_Z2mBQgE.jpg "@1 → the same table, pancakes and mountains, now with the characters. Nano Banana Pro")

To keep the model from redrawing the whole background, say so directly: “keep the cabin, the table and the mountain view.” More scenarios — with a pet, a wedding and a family shot — are in our guide on [how to combine photos](/blog/obedinit-foto-nejroset).

## Characters from “+” and the Builder

The Builder still works with its own character layer, background, clothing and pose. If a character is attached via “+” or the text contains at least one @ tag, AI2Face combines all inputs into one master prompt:

- **the Builder's character** goes first in the list and is also available as an @Name tag;
- **the background, clothing, pose and scene reference** go to the model along with the characters, and the algorithm explains the role of each layer;
- **duplicates don't count twice:** if the same person is selected in both the Builder and “+”, they take one slot.

With no characters in “+” and no tags, the Builder works exactly as before.

## Multiple characters in video

On the Video tab, you attach characters the same way — via “+” → Characters, with the References base. This works in video models that accept reference images, tags included: @Emma points the model to her photo. The example below was made in [Gemini Omni](/models/gemini-omni).

![The Video tab: Gemini Omni 1.1 Flash, characters Emma and Leo, a prompt with @Emma and @Leo tags](https://ai2face.com/_astro/screen-video-omni.Ch2M4U4B_qzqsw.jpg "Video in Gemini Omni: the same characters and tags as in photos")

[A red-haired woman hands a bearded man a small gift at a café table; he finds a film camera and takes her photo](https://ai2face.com/_astro/media/blog/neskolko-personazhej-v-odnom-kadre/gift.mp4 "Gemini Omni 1.1 Flash, references mode: characters Emma and Leo, 16:9, 8 s, with sound")

Omni carried out every action in order: the gift, the film camera, the laugh and the photo of Emma. Both faces match the character photos.

Things to keep in mind for video:

- **The Frames base** doesn't take characters: a frame is a specific photo, not a person. Attached characters come back when you switch to References.
- **Video has a lower limit:** Gemini Omni accepts up to 6 photos, [Veo 3.1](/models/veo-3-1) up to 3. The counter above the prompt shows the limit of the selected model.

## How many characters you can use

A character always takes one slot in the photo limit, even if it has four photos. How many people the model keeps recognizable depends on the model itself.

| Model | Photos per request in AI2Face | Recognizable people (per the developer) |
|---|---|---|
| [Nano Banana Pro](/models/nano-banana-pro) | up to 14 | up to 5 |
| [Nano Banana 2.1](/models/nano-banana-2-1) | up to 14 | up to 4 |
| [GPT Image 2.5](/models/gpt-image-2-5) | up to 16 | not disclosed |
| Gemini Omni 1.1 Flash (video) | up to 6 photos and 1 clip | not disclosed |
| Veo 3.1 (video) | up to 3 | not disclosed |

## One more, Edit and Retry

An image with characters repeats with the same people and tags: One more in the viewer runs it again, and Edit brings the characters, photos and text back into the composer. Thumbnail order is restored so that @1 and @2 point to the same photos as during generation. For a failed job, Retry sends all inputs, including characters and Builder layers.

Filtering the feed and gallery by character finds an image by the first person attached.

## Bottom line

Multiple characters in one image takes three steps: create a character for each person, attach them via “+” → Characters, and name them in the prompt with @Name tags. For photos of people, use Nano Banana Pro; for video, use a model that accepts reference images, such as Gemini Omni. If the faces don't look right, generate 2–4 variants and avoid changing clothing, angle and hairstyle in one prompt — our guide on [how to combine photos with AI](/blog/obedinit-foto-nejroset) breaks down the misses and compares models. For more ways to work with finished shots, see the article on [photo editing](/blog/izmenit-foto-nejroset).

## FAQ

### How many characters can I put in one image?

As many as the model accepts photos: up to 14 for Nano Banana and up to 16 for GPT Image 2.5, with each character taking one slot. Google says Nano Banana Pro keeps up to five people recognizable and Nano Banana 2.1 up to four.

### Do I have to use @ tags?

No, but they make results more accurate. Without tags, the model decides on its own who is who. With tags, AI2Face's algorithm links each name to its photo precisely, so the model puts the right face in the right role.

### Can a character name use any language?

Yes. The tag repeats the character's name: @Pierre, @Даша, @さくら. Tags aren't case-sensitive, and spaces in a name become “_”.

### Do characters from “+” work together with the Builder?

Yes. The Builder's character goes first in the list, and the background, clothing, pose and scene reference go to the model along with it. You don't need a separate Builder layer for a second person.

### Do characters and tags work in video?

Yes, in video models that accept reference images: attach characters via “+” with the References base, and an @Name tag tells the model which of them does what. This article includes a Gemini Omni clip as an example.

## Sources

- https://blog.google/innovation-and-ai/products/nano-banana-pro/
- https://ai.google.dev/gemini-api/docs/image-generation
- https://deepmind.google/models/model-cards/nano-banana-2-1/