Where the style is decided
In two places, and they pull against each other if you are careless. The character itself fixes ethnicity, age, hairstyle, eye colour and body type in the builder. The prompt then decides how that character is rendered. Lock the first, vary the second, and an illustrated look stays recognisable from one batch to the next.
Prompt wording that holds
A prompt can name outfit, pose, action, hair or eye colour, accessories and setting. Fill in more of those six and the result lands closer to what you pictured. Keep the styling words in a fixed order at the front of every prompt and vary only the scene behind them; that habit is what separates a consistent girlfriend or boyfriend from a stranger who shares a name.
Batch sizes and tokens
You can ask for between one and 16 images per request. Every render draws on the 100 tokens included each month, so the cheapest way to work is to prove a prompt on one image, adjust, and only then order sixteen. Tokens can be topped up, do not transfer between accounts, and expire when the billing period ends.
Check every render
Generated images and video can be inaccurate, incomplete or different from what you asked for. Access is for personal use only.
Five-second clips
The video beta takes one generated image and animates it: you pick a scene, a companion, and a movement or action, and the clip runs five seconds. Treat it as a gesture rather than a scene. Generate one during the seven-day free trial so you know what it does, then let the choice of billing term rest on chat and images instead.
Test a style this week
Seven days is enough to find the prompt wording that works.
Opens candy.ai. Adults only.
Questions
How is an anime look set on a companion?
Through the character and the prompt. The builder fixes the features; the prompt carries the styling words on every render.
How many images can one prompt return?
One to 16. Each render spends part of the monthly 100 tokens, so an untested batch of sixteen is the expensive mistake.
Can an anime companion appear in a video?
Yes, through the beta: one generated image, a scene, a movement, and a clip that runs five seconds.
Do I need a plan to generate images?
Yes. Free users get five messages; image and video generation sits behind a paid plan, after the seven-day trial.