Model Type
Prompt techniques and examples for generating images with custom models in Adobe Firefly.
Custom Models is currently in beta for individual and teams users in Adobe Firefly. Try it now and share your feedback.
When generating images with Firefly Custom Models, effective prompts focus on describing the specific visual you want to create rather than the style itself. Your custom model already understands your aesthetic from training, so prompts should specify content details like subjects, poses, settings, and compositions.
Key elements for effective prompts
Strong prompts for custom models include these components:
- Clear subject description: Specify the main subject, object, or character you want generated
- Pose and action details: Describe how subjects are positioned or what they're doing
- Setting and environment: Include relevant background, location, or context details
- Related concepts: Use ideas and terms that align with your training dataset
- Caption-aligned language: When possible, use similar words and phrases from your image captions (for example, use 'ceramic mug' instead of 'cup')
- Concept ID for characters: Include the unique concept ID established during training when generating character-based images
- Specific materials and textures: Add details about surfaces, fabrics, or finishes to refine results
Small adjustments in descriptive language can significantly change outputs. Prompts requesting content unrelated to your training dataset may produce unsatisfactory results—for instance, prompting a model trained on colorful lifestyle photography for black-and-white illustrated spaceships.
Prompt examples by model type
|
|
Example Prompts |
|
Illustration |
|
|
Character |
|
|
Iconography |
|
|
Photography |
|
|
Still life photography |
|
|
Lifestyle photography |
|
|
Photoshoot of a person |
|
|
Isometric and 3D graphics |
|
|
Explore brand expression |
|
|
Backgrounds for product shots |
|
Turn text into art with Adobe Firefly
Create beautiful images and video clips from text prompts using generative AI.