Prompt writing best practices
Be specific and detailed
Good prompt:
A serene mountain landscape at sunset, snow-capped peaks in the background, a calm alpine lake in the foreground reflecting orange and pink sky, dramatic clouds, golden hour lighting, photorealistic, high detail, 4K quality
Include subject, style, composition, lighting, colors, and quality modifiers for best results.
Use style descriptors
Add style information to guide the aesthetic:
- Artistic styles: “watercolor”, “digital art”, “photorealistic”, “sketch”
- Era/period: “medieval”, “futuristic”, “vintage 1950s”
- Mood: “serene”, “dramatic”, “playful”, “mysterious”
Describe composition
Tell the AI how to frame the image:
- Shot type: “close-up”, “wide shot”, “portrait”, “landscape view”
- Angle: “bird’s eye view”, “low angle”, “eye level”
- Focus: “shallow depth of field”, “everything in focus”
Model selection tips
Match model to use case
- Photorealistic images: Use Flux
- Text in images: Use Ideogram
- Artistic illustrations: Experiment with different models
- Specific styles: Try models known for that style
Don’t always use the same model. Different models excel at different things.
Experiment with models
- Try multiple models: Use the same prompt with different models
- Compare results: See which model produces the best outcome
- Learn preferences: Discover which models work best for your needs
Settings optimization
Start with defaults
- Begin simple: Use default settings for your first attempts
- Adjust gradually: Change one setting at a time to understand effects
- Document what works: Keep notes on successful setting combinations
Quality vs speed
- Balance: Find the right balance between quality and generation time
- Higher quality: Use when detail is important
- Faster settings: Use for quick iterations and experiments
Reference image techniques
Choosing good references
- High quality: Use clear, well-lit reference images
- Relevant content: Choose references that match your goal
- Simple compositions: Uncluttered references work best
Combining references and prompts
- Be specific: Describe what to keep from reference and what to change
- Add details: Use prompt to add elements not in reference
- Mention the reference: Explicitly reference elements to preserve
Workflow optimization
Session organization
- Create sessions early: Set up session structure before generating
- Use descriptive names: Name sessions clearly for easy finding
- Organize with folders: Group related sessions together
- Regular cleanup: Archive or delete old sessions periodically
Iteration strategy
- Start broad: Begin with general concepts
- Refine gradually: Add detail in subsequent generations
- Document successes: Keep notes on what works
- Learn from failures: Understand why some attempts don’t work
Cost management
Optimize for efficiency
- Use appropriate models: Choose cost-effective models when quality allows
- Balance settings: Don’t always use highest quality if not needed
- Plan usage: Estimate costs for projects
- Monitor spending: Check dashboard regularly
Budget planning
- Set limits: Establish spending limits
- Track usage: Monitor costs frequently
- Adjust as needed: Modify usage if costs are high
Quality improvement
Prompt refinement
- Iterate on prompts: Refine prompts based on results
- Learn from examples: Study successful prompts
- Use Prompt Assistant: Get AI-powered suggestions
- Combine techniques: Merge different prompt writing approaches
Settings fine-tuning
- Experiment systematically: Change one setting at a time
- Test combinations: Try different setting combinations
- Document results: Keep notes on what works
- Reuse successful settings: Apply what worked before
Common mistakes to avoid
Too vague: Vague prompts produce poor results. Be specific.
Contradictory terms: Avoid mixing conflicting styles or concepts.
Ignoring composition: Without composition guidance, results may not match your vision.
Not experimenting: Always using the same approach limits discovery.
Advanced techniques
Multi-model generation
- Compare results: Generate with multiple models simultaneously
- Discover preferences: See which models work best for different needs
- Save time: Get multiple options in one generation cycle
Prompt templates
Create templates for common use cases:
- Portrait template:
[Subject] portrait, [style], [lighting], [mood], [background]
- Landscape template:
[Location] landscape, [time], [weather], [composition], [colors]
- Product template:
[Product] on [surface], [lighting], [background], [style]
Learning and improvement
Track what works
- Keep notes: Document successful prompts and settings
- Review history: Learn from past generations
- Identify patterns: Notice what consistently works well
- Build knowledge: Accumulate experience over time
Continuous improvement
- Experiment regularly: Try new approaches and techniques
- Stay updated: Learn about new features and models
- Share knowledge: Learn from community and examples
- Practice: Regular use improves results
