Every generation starts with identity. You get a neutral profile headshot and one combined character reference sheet with eight labelled views, the same turnaround animators, illustrators, and 3D modellers ask for before a single scene is drawn.
Wide shots, tight close-ups, new clothes, and even the back of the head are where ordinary image generators drift. A saved Character carries both references into every request, so the brows, the cheekbones, and the proportions survive the change.
Three steps, one saved identity. No fine-tuning, no LoRA training, no re-uploading references for every image.
Write a brief with the details that make the character recognisable: age, build, hair, skin, markings, wardrobe. Or add up to five photos of the same fictional person, pet, or original creature from different angles; the saved Character must not represent a real person. The uploads guide identity and design; they are never copied as scenes.
One generation returns two images with fixed roles: a neutral profile headshot and the combined eight-view sheet. Both are private identity documentation, rendered in flat studio light so nothing hides in shadow.
Save the pair as a private Character. In Create, select it, describe the new scene, the outfit you want in it, and the style, then generate. Portrait requests lean on the headshot; full-body, back, hand, and wardrobe requests lean on the sheet.
CharacterIdentity is separate from rendering. Ask for photoreal, gouache illustration, anime, or a 3D toy look and the defining features carry over, so one character can front a photo campaign, a picture book, and an animated short.
A character generator is only useful if the character survives page two. These examples were made with the tool: each starts from a text brief, gets its sheet, and is then sent into new scenes with nothing but a scene description.
Self-published authors lose readers when the hero's face, raincoat colour, or size changes between spreads. Lock the design once, then illustrate forty pages with the same fox kit, the same yellow raincoat, and the same white sock on the same paw.
Long-running series drift slowly: the jaw softens, the jacket loses its lettering, the profile stops matching the front. A sheet with matching profile and back views keeps dialogue two-shots, action panels, and rain-soaked exits on the same character.
Studios ask freelancers for front, side, and back views before anything goes to 3D. Generate the turnaround and the dialogue bust in one style, then drop the same NPC into the village, the forge, or a round token for the virtual tabletop.
Onboarding screens, empty states, launch banners, and holiday posts all need the same creature with the same colours and proportions. Design the mascot once and generate every seasonal variant without it slipping off-model.
Generic generators return a generic corgi. Describe or upload your pet and the blaze, the folded ear, and the build stay put, whether the dog is on the moon, in a Renaissance ruff, or in a watercolour on the beach.
Upload up to five photos of the same fictional person, pet, or original creature and the generator uses them together as identity and design guidance. No training, no LoRA, no waiting.
A front view, a profile, a full-body shot, and a back view each reveal something the others hide. Five near-identical selfies do not.
Even daylight and an unobstructed face give the model the detail it needs. Sunglasses, heavy filters, and motion blur remove it.
Name the age, build, skin, hair, distinguishing marks, and wardrobe. Specific, imperfect details make a more consistent character than "beautiful".
Most "consistent character" features attach one face image to a prompt and hope. The difference here is what gets saved and what gets sent.
| Capability | Typeshutter Character | Single-image reference | Prompt only |
|---|---|---|---|
| Identity source | Text or up to five photos | One image | Re-typed description |
| What you get | Headshot plus eight-view sheet | One image per prompt | One image per prompt |
| Profile and back views | Documented on the sheet | Guessed | Guessed |
| Reuse in new scenes | Select the saved Character | Re-attach the image every time | Re-describe every time |
| Training required | None | None | None |
Recent public posts from the community, created with this tool.
Generate the character once, save the headshot and reference sheet as a Character, then select that Character in Create for every new image. Describe the scene, the outfit, the pose, the lighting, or the style you want in that image; the identity is not re-described. The saved references are sent with each request, so the face and body stay stable instead of being re-imagined from the prompt.
Ordinary generators have no memory. Every prompt re-solves the face from scratch, so features drift with each scene, and they drift most in wide shots, extreme angles, and strong expressions. A saved Character fixes this by attaching a neutral headshot and an eight-view sheet to each generation, giving the model documented angles to match rather than a description to reinterpret.
Yes. With GPT Image 2 you can upload up to five photos of the same fictional person, pet, or original subject; Grok Imagine Image 2.0 accepts up to three. One clear photo works, but complementary views such as front, profile, full-body, and back produce a more complete sheet than several near-duplicates. Photos guide identity and design only; they are not reproduced as scenes.
New accounts receive trial credits, and a character generation is priced as its two output images at standard per-image credit rates. You can save one Character on a free account and more on a subscription.
You can use what you generate in your own projects under the Typeshutter terms. Characters must be fictional: when you save one, you confirm it does not represent a real person. Check the rules of the marketplace you publish to, such as disclosure requirements for AI-assisted books.
A character reference sheet, also called a model sheet or turnaround, documents one character from the angles an artist or animator needs to stay on-model. The sheet here contains four neutral face angles (front, 3/4 left, 3/4 right, left profile), three complete standing views (body front, body side, body back) with hands and feet, and a shoulders-to-hips torso crop, all with readable labels on a neutral studio background.
Yes. Select the saved Character and describe the pose, outfit, or expression you want. The sheet's neutral expressions do not limit future ones. Clothing is not copied from the saved references automatically, so name the wardrobe in each scene request whenever it matters; the face, body, and markings carry over regardless.
Any style the selected model can render. The character's identity is separate from the rendering, so the same Character can appear as a documentary photograph, a gouache picture-book illustration, a cel-shaded webtoon panel, or a 3D toy render while keeping its defining features.
Yes. The generator accepts people, pets, mascots, and original creatures. Animals and naturally covered creatures keep their own anatomy and covering on the sheet, and markings such as a blaze or a folded ear are documented so they persist in later scenes.
Generate a scene with the saved Character, then open it in the AI image to video tool. The scene becomes the opening frame, so the identity you locked carries into motion.
Those features attach a single image to a single prompt and must be re-attached for every generation. Here the character is a saved object with two purpose-built references, including documented profile and back views, that the app sends automatically with every request. No fine-tuning or LoRA training is involved.
Yes. The headshot and the combined sheet are ordinary images in your library. Download the sheet at 2K, or at 4K when it was generated with GPT Image 2 (Grok Imagine Image 2.0 tops out at 2K), and hand it to an animator, rigger, or image-to-3D pipeline as you would any turnaround.
One or more selected files can’t be uploaded. Please upload different files, or change the format.
