Get started

Phota Studio works from a profile you train on your own photos. Once that profile exists, every edit and generation keeps you looking like you. This guide walks through training the profile, then using the three tabs it unlocks.

Open Phota Studio See pricing

Train your first profile

A profile is a private model of one person or pet. Nothing else in Studio works until you have one.

Training a profile end to end, from the People page through to the training run.

1.
Open the People page. It lists every profile on your account, and it is where training starts.
2.
Press Add. The Add button at the bottom of the page opens a new, empty profile. Find People beside it does something different: it searches photos you have already uploaded.
3.
Name them. Type the person or pet’s name into the Name field. This is the name you will tag them by later, so keep it short and distinct from your other profiles.
4.
Add their photos. Drag files straight onto the panel, or use Upload from Google Photos or Upload from Phone. The counter tracks how many have been accepted. Which photos you pick matters as much as how many: what makes a good training set.
5.
Pick a training mode. Full Train or Quick Train, using the toggle above the counter. The one you pick changes how many photos the panel accepts, which is why the counter target moves with it.
6.
Press Create. Training spends one subject token and takes about fifteen minutes. The status moves through Validating, then Queuing, then Training. You can leave the page while it runs.

Quick Train or Full Train

Quick Train: 5 to 10 photos. The faster turnaround, and enough to try Studio out on someone you have few photos of.
Full Train: 10 to 50 photos, and the mode to use for anyone you will generate with regularly. Quality keeps improving up to 50, so aim for the full set rather than the minimum.

What makes a good training set

Two things decide how good a profile turns out: how many photos you give it, and how different they are from each other. Fifty near-identical frames from one afternoon teach the model one lighting setup and one expression. Twenty photos across years, places, lighting, and moods teach it what you look like.

A training set worth aiming for: faces you can read, shot from different angles, in different moods, some of them alongside other people.

Clear, visible faces: The face should be readable: in focus, lit well enough to see, and big enough in the frame. Vary the light across the set. Daylight, indoor light, evening, and shade all help.
Angle and pose variety: Straight on, three-quarter, profile. Looking at the camera and away from it. Close-up portraits, upper body, and a few full-length shots.
Expression and accessory variety: Different expressions, hairstyles, hats, sunglasses, different clothing. Photos from different years are good, not a problem.
Group shots are fine: You do not need solo photos. A shot with other people in it works, as long as the face you are training is big and clear enough to read.
Leave out: Near-duplicates from a single burst, heavy filters, and anything low resolution or motion-blurred.

Booth, Editor, and Actions

Three tabs, split by what you start from: a prompt, a photo, or a one-click action.

Booth: Generation. Describe the photo you wish you had and Studio makes it from scratch. Tag your people to put them in it.
Editor: Editing. Bring a photo you already have, describe the change, and Studio applies it. You can attach up to ten images at once.
Actions: Quick actions. The four most common edits, each one click, with no prompt to write.

Tag people with @mentions

A tag is how you tell Studio which trained profile a prompt is talking about. Type @ anywhere in a prompt and pick a profile from the list that appears. The tag highlights in yellow once it matches. Tags work in Booth prompts, Editor instructions, and the optional instructions on a quick action.

Without a tag, Studio has no way to know which of your people you mean, and will render a generic face instead. A tag that does not match a profile shows a question-mark badge; click it to correct the name.

Typing @ in the prompt box opens the list of trained profiles.

The four quick actions

Fix it

Restores clarity and detail in a photo that came out soft, noisy, or badly lit.

Pro look

Polishes an ordinary photo into a professional-looking shot.

Add people

Adds people and pets to a photo they were not in. Pick who from your trained profiles.

Profile to add

Style me

Applies the look of a reference photo you supply to your photo.

Every quick action takes optional instructions — here, telling Style me to keep the result photographic rather than painterly.

Name the faces in a photo

When you bring your own photo in, Studio finds the faces. Telling it who is who is what keeps the right identity through an edit.

Detected faces appear as circles on the photo. Any face that hasn’t been assigned to a profile is marked Unknown. Click a face to open the assignment popover, then choose a profile from the list or search by name.

You can change an assignment at any time by clicking the face again. If Studio guessed wrong, correct it here. Rewording the prompt won’t fix an incorrect assignment.

An unknown face will still render, but the model won’t know whose identity to preserve. For anyone who appears regularly in your photos, train a profile for them and assign it here. That’s what keeps them looking like themselves through an edit.

Assigning a detected face, and what to do with one that has no profile yet.

Choosing a base model

Phota is model agnostic. Your profile works across every base model Studio offers, so which one you pick is a question of the look you want.

Studio offers four base models. Nano Banana 2 is the default and the safest starting point. Switch when you want a different look, and expect the settings to move with it: a model that does not support 4K or the Auto aspect ratio drops those options, and the picker snaps to a supported value.

The model you pick does not change what an image costs. Studio bills by resolution alone: 2 credits at 1K, 3 at 2K, 4 at 4K.

Model	Resolutions	Auto aspect ratio
Nano Banana 2	1K, 2K, 4K	Yes
GPT Image 2	1K, 2K, 4K	Yes
Seedream 5.0 Pro	1K, 2K	Yes
Flux 2 Pro	1K, 2K	No

Nano Banana 2 is the default. Fixed aspect ratios (1:1, 3:4, 4:3, 9:16, 16:9) are available on every model. Other models exist behind the Phota API but are not offered in Studio.

Ask for a feature

Tell us what is missing, or what is not working.

Send feature requests, bugs, and questions to support@photalabs.com. If it is about a specific result, include the photo and the prompt you used.