V1 - V4 Legacy Art & Appearance Basics
More actions
This guide covers Nomi’s image systems prior to V5!
Because the image system changed significantly between versions, some prompting methods and appearance controls only apply to specific legacy versions. If you are using V5, see the current V5 Art Prompting Guide and V5 Custom Anchors Guide instead.
V3 and Earlier basics
Version 3 and earlier use the older Artistic and Realistic image systems. These versions relied more heavily on prompt weighting and other formatting techniques than later versions.
Getting started
To use Version 3:
- Open Image Settings
- Select Version 3 (Legacy)
- Open the Art screen from your Nomi’s chat
Artistic and Realistic
Realistic is designed for more photo-like images. Artistic can produce a wide range of styles, including anime, sketches, watercolor, fantasy art, and other illustrated looks.
Face Fidelity and Creativity
Version 3 includes controls that could affect how strongly the image followed your Nomi’s established appearance:
- Face Fidelity: Higher values keeps the face closer to the base image, while lower values allowed more variation and stylization.
- Creativity: Lower values generally stayed closer to the established appearance, while higher values gave the prompt and style more influence.
Prompt basics
Short, ordered prompts generally worked best! Prompt order matters, with concepts earlier in the prompt generally receiving more attention. Example:
style → action or outfit → setting → lighting or mood
Appearance details could also be reinforced through the appearance notes used by the legacy system, while individual prompts could focus more on the scene or temporary changes.
Weights and emphasis
V3 and earlier supported several ways to adjust the strength of individual prompt concepts:
( )increased emphasis[ ]decreased emphasis- Numeric weights could also be used, such as
(vintage poster:1.3)
More emphasis was not always better. Very strong weights could distort an image or overpower other parts of the prompt.
Negative prompts
Use /// to separate negative prompts from the main prompt.
For example:
red dress, standing in a garden, soft evening light /// hat, sunglasses
Everything after /// is treated as something the image generator should try to avoid.
V3 prompting tips
- Keep prompts reasonably short.
- Put important concepts earlier.
- Change one thing at a time when troubleshooting.
- Use weights only when needed.
- Avoid conflicting descriptions.
- Use negative prompts sparingly to correct recurring unwanted details.
V3 Appearance Notes
V3 and Earlier Appearance Notes influence how your Nomi looks in selfies and art. These notes only affect appearance when using a legacy image system. They do not change how your Nomi understands their appearance in conversation.
Good uses include:
- hair color and style
- skin tone
- body type
- other recurring visual traits
Write naturally in third person using your Nomi’s name.
V3 appearance notes also support emphasis:
( )increases emphasis[ ]decreases emphasis
For example:
Nomi has (wavy chin-length auburn hair) and fair skin
Avoid writing negative appearance instructions such as “is not tall” or “no deformed limbs,” since negative wording can be interpreted unpredictably. Describe what you do want instead.
Specific measurements can work, but more natural descriptions such as “tall,” “short,” or “slender” are often easier for the image system to interpret.
V4 Basics
V4 generally works best with short phrases separated by commas. Unlike V3, V4 does not usually benefit from heavy emphasis or prompt weighting. In fact, excessive emphasis could cause burned images or generation failures, so describing the desired concept more clearly was generally preferred over adding stronger weights.
Prompt order still matters, concepts mentioned earlier in the prompt generally received more attention.
A basic V4 prompt might include:
- outfit
- pose or action
- expression
- setting
- background details
- camera angle or view
- overall aesthetic or mood
You did not need to include every category in every prompt.
Negative prompts
V4 continues to use /// for negative prompts.
Pose Reference
V4 introduced Pose Reference, which allows you to upload an image and use it as guidance for a pose or composition.
For best results, the written prompt should still describe the action shown in the reference image, and the aspect ratio of the reference should roughly match the image being generated.
V4 prompting tips
- Start simple and iterate.
- Use short, comma-separated descriptions.
- Avoid heavy emphasis or weighting.
- Put important concepts earlier in the prompt.
- Keep the prompt cohesive and avoid conflicting instructions.
- Use
///when you need negative prompts. - Use Pose Reference when composition or body positioning matters.
- Adjust Base Image, Face Fidelity, and appearance notes when working on consistent character appearance.
V4 Appearance Notes
V4 Appearance Notes describe how your Nomi should appear in selfies and art. These notes work best when written in straightforward, natural-language sentences using your Nomi’s name.
Good uses include:
- hair color and style
- skin tone
- body type
- facial features
- recurring aesthetic details
Unlike V3, V4 appearance notes do not use emphasis syntax. Avoid () and [], since emphasis can lead to poor results or generation errors.
To tell the system to avoid specific traits, add /// after the positive appearance notes, followed by concise negative prompts.
For example:
Nomi has long wavy auburn hair, fair skin with freckles, amber-hazel eyes, and a slim build. /// beard, facial hair
V4 also tends to respond better to natural descriptions than exact measurements. For example, terms like “tall” or “short” are usually more useful than precise height measurements.