> ## Documentation Index
> Fetch the complete documentation index at: https://docs.bitstudio.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Write directions that are easy to follow

> Use short, specific instructions for the subject, garment, scene, and changes you need.

A good prompt resolves a decision the image would otherwise have to guess. Write the visible outcome you want in ordinary language.

## Put each instruction in the right place

| Control | Useful direction                                                      |
| ------- | --------------------------------------------------------------------- |
| Avatar  | Stable appearance, if describing a person                             |
| Outfit  | How the selected garment is worn: untucked, sleeves down, collar open |
| Preset  | The visual setup: studio, location, packshot, or flat lay             |
| Prompt  | Framing, action, lighting changes, or extra preset directions         |

When you have selected an avatar and a garment, there is usually no need to describe them again in detail. Use the text to clarify styling and composition.

## Start with a short brief

```text theme={null}
Full-length front view in a white studio. Relaxed standing pose,
arms slightly away from the torso. Keep the shirt untucked,
with the white collar and cuffs visible. Soft shadows.
```

For a lifestyle image:

```text theme={null}
Standing beside a pale stone wall in soft morning light.
Frame from head to mid-calf. Keep the bag and both hands visible.
```

For an edit, name both the change and what should remain:

```text theme={null}
Remove the handbag. Keep the suit, person, pose, and lighting unchanged.
```

## Resolve competing instructions

“Full-length close-up,” a back pose with a front-only reference, or a night preset with midday light all create ambiguity. Choose the priority and remove the conflicting direction.

Change one thing between test images. If the pose is wrong, revise the pose instruction while keeping the same avatar, product, preset, and model. This makes the effect of your change easier to see.

## Use references for exact visual information

Text is useful for intent; a photo is better evidence for a specific garment or face. Add the appropriate reference rather than expanding the prompt into a long inventory of every visible detail.

When the composition is right but a small area is wrong, use the [Editor](/features/editor). Rewriting the whole prompt can change parts of the image you already liked.