Character Studio
Create a brand new face, character, or object. Design unique AI identities from scratch. From 15 credits an image.
How do I create a fictional AI character from scratch?
Character Studio builds a face from controls rather than from photographs. You set 8 attributes — gender, style, hair style, hair color, eye color, skin tone, face shape and expression — or write a description of your own, and each run returns one image of a person who does not exist. Nothing is trained and nothing is uploaded. To bring that same face back, carry the picture you keep into a tool that works from references.
- Credits
- 15–20 an image, 15 on the engine and size the studio opens with
- Input
- None required — controls, or a sentence
- Per run
- One image, at the aspect and size you choose
- Subject
- A person who does not exist
Updated September 4, 2026
Cite this page: https://imagera.ai/image/personal-influencer/create-character
What an invented face is good for
A new person, not a copy of one
It's worth being clear about the boundary. Character Studio is for inventing a new person, not recreating someone specific. If you want images of yourself or of cleared talent, Imagera's face-training path handles that. Character Studio starts from attributes and imagination, which is exactly what makes it fast for personas that need to exist independently of any real individual.
Know more →For work with no model to photograph
The tool is built for people who need a consistent face but don't have, or don't want to use, a real model. Virtual influencer creators use it to run a believable social persona without a photoshoot. Brands build a spokesperson or mascot that shows up the same way across a campaign. Game studios, comic artists, and app teams use it to lock a recurring character early, before any casting or 3D work. Because the person is invented, there is no model release to sign, no scheduling, and no likeness rights to clear — the friction that usually surrounds using a real face simply isn't there.
Know more →A short form, not a prompt-engineering exercise
It's also genuinely accessible. There's no GPU, no install, and no prompt-engineering requirement — the controls are a short form, and you can go from a blank slate to your character's first images in a couple of minutes. Pricing is credit-based, from 15 credits per image, with no per-character subscription and no per-seat fee, so the cost scales with how much you actually generate rather than with how many personas you keep on file.
Know more →How a face gets designed
8 controls, or a sentence of your own, and one picture back per run.
The controls are a casting brief
The card offers 8 attributes — gender, style, hair style, hair color, eye color, skin tone, face shape and expression — each of which can be left on Any. Whatever you set is composed into the description the run is built from, so you can arrive with nothing in mind and still leave with a specific person. Typing your own description overrides the controls completely: the composer only assembles a sentence when you have not written one, which means the short form is a starting point rather than a cage. A run with neither a description nor a single attribute is refused rather than guessed at.
Try it now →One picture per run, and what it costs
Each run returns a single image at the aspect ratio and output size you pick, and the size is the biggest lever on the price: 15 credits on the engine and size the studio opens with, 15 on the higher-quality engine at that same size, and up to 20 at the largest size either offers. There is a second mode that edits a picture you upload instead of creating from nothing, at 15 credits a run. Every combination is priced on the button before you commit, and it comes out of the same credit balance as the rest of the suite.
Try it now →Making the face come back is a second step
Nothing here is saved as an identity, and that is worth knowing before you plan a feed around it: a run generates a face, it does not enrol one. The picture you keep is what carries the character forward. Drop it into the optional reference slot on this same card to steer the next generation towards it, or hand it to Instant Character, which is built to hold one face across many scenes from reference pictures. For a character that has to survive months of content, the picture becomes the first photograph of a training set instead.
Try it now →How to create a fictional AI character
Four steps in the browser. No photographs, no install, no GPU.
Set the attributes, or describe the face
Work down the 8 controls — gender, style, hair style, hair color, eye color, skin tone, face shape and expression — leaving anything you have no opinion about on Any, or skip them and write the description yourself. A description you type wins over the controls, so you can start with the form and take over whenever you want to.
Pick the frame shape and the size
Choose an aspect ratio and an output size. The size is what moves the price: the studio opens at 15 credits and the largest option on the higher-quality engine is 20.
Generate, and generate again
Each run returns one image. Nothing about the face is fixed between runs, so regenerating is how you audition it — keep going until the eyes, jaw and proportions read as one specific person rather than a type.
Keep the picture you want to build on
The image you save is the character. Feed it back into the reference slot here, or into Instant Character, when you want that same face in a new scene — and into a training set when it needs to hold up over a long run of content.
Who needs a face that belongs to nobody
A brand face with nobody to book
A campaign that wants a human face but has no budget for talent, no booking calendar and no release to chase can start with one that was never a person. The face is designed rather than cast, so it costs a generation instead of a shoot day, and it can be revised on the spot when the brief moves — which is the part a real casting cannot do at all.
Try it now →A character that has to exist before anyone is cast
Game, comic, animation and product teams routinely need to see a character before any of the expensive work starts. A designed face turns a written brief into something a room can react to in minutes, which makes it a cheap way to settle an argument about a look — and if the project later needs a real model or a 3D build, you go into that conversation with a picture rather than adjectives.
Try it now →The starting photograph for an identity you will train
A trained model needs photographs of its subject, which is awkward when the subject has never been photographed. Designing the face is how that gets solved: the picture you keep here becomes the reference a training set is built from, and from that point an invented person takes exactly the same route through the suite as a real one.
Try it now →Discover more AI features
Pro LoRA Generator
For advanced users — bring your own LoRA or choose from 1,000+ community styles and characters.
Lipsync Studio
No camera, no shoot, no retakes — a photo and a voice note become a video of that person saying it, or a clip you already shot says something new.
AI Influencer Generator
Create a consistent virtual influencer — generate a face or train yours, then make photos and reels
LoRA Trainer
Train custom AI models on your own images
Frequently asked questions
What does Character Studio actually produce?
One image per run, of a person who does not exist. You set the attributes — gender, style, hair style, hair color, eye color, skin tone, face shape and expression — or write your own description, and the run returns a portrait built from it. Nothing is uploaded and nothing is trained. A second mode edits a picture you supply instead of creating from nothing, but the default is genuinely from-scratch: the only input is the brief.
Does the same face come back on the next run?
Not on its own. There is no save step and no stored character — each run is a fresh generation, so the same settings will give you a person of the same description rather than the identical individual. Making a face recur is a step you take with the picture: put it in the optional reference slot here to steer the next run towards it, or use Instant Character, which is built to carry one face across many scenes from reference pictures. For a character that has to stay identical across months of content, the picture becomes the first photograph of a training set.
What can I actually control about the face?
8 attributes: gender, style, hair style, hair color, eye color, skin tone, face shape and expression. Each can be left on Any, and anything you type as a description overrides all of them. There is no age control and no ethnicity control on this card, whatever else you may have read about it — those two settings do not exist, and asking for them in your own wording is the only way to reach them.
What does one image cost?
15 credits on the engine and output size the studio opens with. The higher-quality engine is 15 at that same size, and the largest size either engine offers is 20; editing an uploaded picture rather than creating one is 15. There is no training run to pay for and no per-character charge — the price is per image, shown on the button before you commit, and drawn from the balance the whole suite shares.
Do I need photographs, or any design skill?
Neither. The default mode takes no picture at all: the controls are a short form, and leaving every one of them on Any and writing a single sentence is a valid run. There is no GPU, no install and no prompt-engineering trick required to get a usable face — regenerating a few times to audition the look is the whole technique.
How many characters can I make?
There is no limit and no per-character charge — you pay per image, so a second character costs exactly what a second image costs. Because nothing is stored as an identity, keeping several apart is a matter of keeping their pictures apart: the file you saved is the character, and it is what you feed forward when you want that one back.
Designing a face vs casting one vs letting a generator invent one
| Dimension | Imagera Character Studio | Casting and photographing a real model | A plain image generator |
|---|---|---|---|
| What you need before you start | A few settings, or one sentence. No photographs and no subject. | A person who fits the brief, a release, a booking and a shoot day | A prompt, and enough practice to write one that lands |
| Who the face belongs to | Nobody — it is generated, so there is no release to sign and no calendar to work around | A real individual, with the rights and scheduling that come with them | Nobody, but you get a different one on almost every prompt |
| How the look is directed | 8 named controls — gender, style, hair style, hair color, eye color, skin tone, face shape and expression — or your own sentence, which overrides them | Casting, styling and art direction, decided on the day | Whatever you can get into the wording |
| Making the same face come back | A second step: carry the picture into the reference slot here, into Instant Character, or into a training set | Book the same person again | Not supported — the face changes with the prompt |
| What one run produces | One image, at the frame shape and output size you chose | A shoot's worth of frames, once, of one real person | One or more images, of one or more strangers |
| Cost | 15–20 credits an image (15 as the studio opens), from the balance the whole suite shares | Talent fees, crew, location and the time to arrange all three | Per-generation cost, and rights that vary by tool |
Use responsibly
People also ask about commercial use. Images generated on a paid plan carry full commercial rights, so a fictional character can appear in social posts, paid ads, storefronts, packaging, and client campaigns. Because the person is invented rather than based on a real individual, you sidestep the likeness-clearance issues that come with a real model — though you should still follow each platform's rules and disclose AI-generated content where it's required.
A designed face is a generated picture, not a person: it has no biography, it has never said or done anything, and presenting it as a real individual — a customer, a testimonial, an employee — misleads the people who see it. Generated faces can also land close to somebody real by coincidence, so if a character is going to be used at scale it is worth looking at it with that in mind before it goes out. The second mode edits a photograph you upload, and everything that applies to the likeness of a real person applies there in full.
A face with nobody to book
Design it from the controls, keep the picture you like, and carry it forward when you want that face to come back.
Design a face →