Writing Image Prompts

Your prompt is the text that tells the image generator what to create. This page shows you how to write one for each model. New to the generator? Start with Image Generation.

Prompt Architect, in the Community tab, can help you draft image and video prompts. Review the prompt before using it.

Each model requires different prompts

Before you try this: Generate one image and note the Model and Model Version that you used.

💡

Power tip: On Experimental Realistic, weights and BREAK give you the most control. Save a plain-language baseline first, then compare one syntax change at a time.

Use these formats as starting points:

💡

Power tip: Change one thing at a time. If you change the prompt and model together, you will not know which change helped.

Getting a better shot

Give the scene a specific place, light source, and action.

Less specific:

standing, looking at the viewer, simple background, soft daylight

More specific:

leaning on a wooden counter in late window light

Name the style. Try live-action, anime, oil paint, comic, or 1970s print.

Adding Weights in your prompt

Power tip

Before you try this: Start with a working prompt on the Base or Companion V2 model.

A weight increases or reduces the emphasis on a phrase. Use it to refine one detail, such as a scarf that you want to stand out.

The V3 model does not use weights. Describe the detail in plain sentences instead.

How to write a weight

Put the phrase in parentheses, followed by a colon and a number:

(red scarf:1.3)

Weights guide the result. They do not guarantee that an item appears or disappears.

Starting numbers

For the Companion V2 model, use these as starting ranges:

Change one weight by 0.1 at a time and compare the result.

High weights can distort details. Keep most of the prompt unweighted.

V3 model works differently

Before you try this: Choose Nectar Companion Model under Model, select your companion, then choose V3 or V3 2K under Model Version.

💡

Power tip: Before changing a Companion V3 setup, save a baseline image and compare one appearance change at a time.

Write in plain sentences, as if you are describing a photo to a friend. Weights and BREAK do not work on the V3 model.

See Adding weights for supported options.

Short prompt: V3 picks the details

Keep the prompt short when you want V3 to choose some of the details.

Mika on a stone rooftop at night. Choose her outfit and pose. Full moon and city lights.

This sets the scene while leaving the outfit and pose open.

Detailed prompt: V3 follows your description

When you want a specific result, describe:

1. Shot or style.

2. Who is in the picture.

3. Clothes.

4. Action or pose.

5. Place and lighting.

6. Viewpoint.

Mika stands on a stone rooftop at night in a shiny black latex catsuit, looking at the viewer. Full moon, city lights. Wide shot.

Describe visible actions. “She leans on the railing and laughs” gives clearer direction than “she feels carefree.”

For group images, give each companion a separate description. See Multiple companions.

Changing a companion’s outfit

Ask for the new outfit in chat and describe it explicitly.

In the full generator, use the prompt for the scene, lighting, and pose. A detail you add to Additional Physical Traits will show up in new images, but removing one does not take it back out, so ask in chat when you want a different outfit.

Multiple companions

Before you try this: Generate a single-companion image first. Choose Nectar Companion Model under Model, then select V3 under Model Version and the companions for your group image.

You can include up to 10 companions. They do not have to be yours. Use Discover in the companion picker to add public companions to the scene.

One companion, one paragraph. Start each paragraph with the name exactly as it appears on the companion’s card. Then describe their clothes, action, and position.

Start with a scene header:

[Shot type or style] in [place], [light]. [Total] people only.

For example, with Sophia, Kirie, and Mika selected:

Wide night shot of a motel parking lot, neon reflected on wet asphalt. Three adults only.

Sophia
Leather jacket and jeans. Standing beside the open driver's door of a parked sedan on the left.

Kirie
White tank top and a gold chain. Standing beside the motel entrance in the center.

Mika
Shorts and a loose shirt. Standing in the foreground on the right, looking toward Kirie.

Give everyone a distinct position before adding complex interactions.

If you want a clearer composition, simplify the scene or start with fewer companions.

Troubleshooting

Try one adjustment at a time.

For clothing, replace a vague instruction:

describe the body, then mention an outfit

With a direct description:

wearing a blue sweater covering the torso and full-length jeans, fully clothed

Next: Video Generation · Customizations · Stuck? Troubleshooting and Getting Help