Experiences
Write an image prompt for your Candy AI character
Decide who appears, what they are doing and how close the frame should be. Add the clothing, background and light that belong to that moment. Our picture requests work best as clear visual descriptions, not as a transcript of everything that happened in the character conversation.
On this page
Choose the camera view before adding small details
A close portrait, a waist-up composition and a wide environmental view prioritise different things. Match the pose to the view you request. A distant figure cannot provide the same facial emphasis as a close-up, and a crowded background can conflict with the simple composition you intended.
The companion selection identifies the character in the supported tool. Use the request to describe this particular frame rather than repeating a complete personality or scene history. Keep one principal activity and add only surroundings that make that activity visually understandable.
Portrait example: preparing a gallery label
This example has a defined frame, one object and a consistent light source. It gives you concrete details to inspect afterward: the character's position, the card and the background. Adapt those details to your own idea rather than combining several unrelated versions into one prompt.
Show the selected adult companion from the waist up at a gallery worktable. Give them a forest-green shirt and a small exhibition label held in one hand. Use soft light from a side window, a calm expression and a lightly blurred wall of empty frames behind them.
Anime example: waiting beside the evening tram
For an illustrated picture, name the style while preserving the same subject-and-frame structure. A simple action and a small number of setting details leave less ambiguity about the moment. Inspect where the object and character appear before changing the whole scene.
Create an adult anime companion at a quiet tram stop just after sunset. Use a rust-red coat, a closed umbrella held at their side and a three-quarter view. Put a route map behind the character and warm station lighting above, with an uncluttered platform.
Revise what you can identify in the output
Use ordinary visual language unless the current interface offers a specialist field. Seeds, samplers, weighting notation and negative-prompt formats from a different generator are not automatically supported. Copying a technical string into a description does not establish that those parameters are active.
Pick the most important discrepancy, remove contradictory wording and make a focused revision. A longer prompt is not necessarily a more coherent one. Check the displayed charge before another generation; repeated attempts do not promise identical identity, a particular success rate or exact control of every detail.
Your questions
Should I add more words after every disappointing picture?
Add a detail only when it clarifies the intended frame. Remove conflicting poses, views or settings first. A short, targeted correction is easier to evaluate than changing all the visual choices at once and then trying to guess which instruction affected the next result.
Are technical tags from other generators supported?
Only rely on a specialist setting when the selected Candy AI tool actually exposes it. Different generators can interpret fields differently. A copied negative-prompt string or seed value does not create a hidden control or guarantee that the same syntax is recognised here.
How should I revise an image with the wrong framing?
Say which part of the view needs to change, such as moving from a full-body frame to a waist-up portrait. Keep the compatible scene details and review the next output. Each new request can have a cost, and no instruction guarantees exact fulfilment.