A good AI avatar prompt names who the character is, how they're framed, the style, and the light, in that order. The harder part comes after the first good result: getting the same face again in the next image. That's what a Character ID is for. It turns a photo or a description into one reference sheet you can reuse in any image or video by typing @ and its name in Sandbox. Below are copy-paste avatar prompts for common styles, then the workflow for keeping one character consistent.
1.The avatar prompt formula
- Identity: age range, build, hair, skin tone, and one detail you'd recognise them by (a scar, freckles, round glasses).
- Framing: head-and-shoulders, waist-up or full body; facing the camera or three-quarter.
- Style: photo, 3D render, anime, flat illustration, pixel art.
- Wardrobe: one or two garments, with colours.
- Light and background: "soft studio light, plain light-grey background" keeps an avatar clean and reusable.
Keep the identity part the same, word for word, every time you prompt the same character. Changing "auburn bob" to "red short hair" is enough to get a different person.
2.Copy-paste AI avatar prompts
Professional headshot avatar
Head-and-shoulders portrait of a woman in her 30s with an auburn bob and freckles, navy blazer over a white shirt, facing the camera with a calm smile, soft studio light, plain light-grey background, realistic photo, 85mm lens.
Friendly creator avatar
Waist-up photo of a man in his 20s with curly black hair and round glasses, mustard hoodie, laughing, bright daylight, pastel blue background, realistic, shallow depth of field.
3D-style avatar
Stylised 3D character render of a teenage girl with a high ponytail and a denim jacket, big expressive eyes, soft rim light, plain gradient background, head-and-shoulders, clean toy-like materials.
Anime avatar
Anime-style portrait of a young man with silver hair and a black high-collar coat, calm expression, three-quarter view, cel shading, soft blue background, clean line art.
Gaming avatar
Head-and-shoulders portrait of a futuristic courier with a shaved head, a thin scar across the left eyebrow and a glowing orange visor pushed up on the forehead, neon city bokeh behind, cinematic rim light.
Fantasy character
Full-body portrait of an elf ranger with long braided dark hair, a green hooded cloak and a longbow, standing in a misty forest clearing, soft morning light, painterly realism.
Sci-fi character
Waist-up portrait of a starship engineer in her 50s with cropped grey hair and grease on her cheek, worn orange flight suit, standing in a dim engine room, practical lights, realistic photo.
Flat illustration avatar
Flat vector illustration avatar of a smiling person with a curly afro and gold hoop earrings, simple shapes, limited palette of teal, coral and cream, circular crop, no background detail.
Pixel-art avatar
64x64 pixel-art avatar of a knight with a red plume and a dented steel helmet, front view, limited 16-colour palette, transparent background.
Mascot for a brand
Friendly cartoon mascot of a round orange fox wearing a small green scarf, waving, simple bold outlines, plain white background, full body.
3.Keep the same character in every image
Prompting alone drifts. Even with identical wording, each new image makes small decisions about the jaw, the eyes and the hairline. A Character ID fixes that by giving every model the same picture to work from.
It isn't a trained model. It's one high-resolution reference sheet: four full-body views (front, three-quarter, side, back) and five expressions, all in the same light. Any model that accepts a picture can use it, including new models as they launch.


Example: "Juniper", a Character ID made from one photo. Four views and five expressions, one face.
3.1How to make and use a Character ID
- Open Character ID and add up to 6 clear photos of the person or character, or describe them in words. Two to five angles work best.
- Name it, for example
juniper-ref. The sheet is created once and saved to your Vault. The credit cost shows before you create it. - Type @ in Sandbox and pick the ID. The sheet joins the inputs of any model that takes a picture.
- Write only the new scene. "Walking through a rainy night market, neon signs, wet pavement." The ID carries the face, hair and outfit.
- Make a video from it. The same sheet works as the reference for image-to-video models, so the face holds in clips too.

Example: made in Sandbox with @juniper-ref as the reference.
The same approach works for things that aren't people: Object ID for a product, Location ID for a place, Outfit ID for a look you want to put on different characters. All four live on the Character ID page.
4.Avatar prompt mistakes to avoid
- Changing the identity words between prompts. Copy them exactly.
- Stacking styles. "Anime, Pixar, oil painting, photoreal" produces a muddle. Pick one.
- Busy backgrounds for a profile picture. A plain background crops and scales cleanly.
- Hands near the face in small avatars. They're the first thing to go wrong.
- Using a stranger's photo. Only make an ID from photos of yourself, people who have agreed, or characters you created.
If a face comes back distorted, see how to fix distorted faces in AI images. For more prompt ideas, browse the AI image prompts guide.
5.What it costs
Character IDs and generations use credits, shown before you run. The Starter plan is $19.99/mo for 400 credits, and credit packs start at $39.99 for 350 credits. See pricing.



