Generation Workflow Overview
Creating a thumbnail involves four key components working together:1
Face Profile Selection
Choose which face profile should appear in the generated thumbnail.
2
Template Selection
Choose a starting template from your library or generate from scratch.
3
Reference Assets
Upload supporting images like logos, products, or screenshots.
4
Prompt Writing
Describe the specific modifications and composition you want.
Step 1: Select Your Face Profile
From the generation interface, select the face profile you want featured in the thumbnail. This determines whose face appears in the final output.Make sure your face profile is properly configured before generating. See the Face Profile Optimization guide for setup instructions.
Step 2: Select a Template
Templates are starting images that the AI uses as a base for your thumbnail. You can import them from YouTube or upload your own.How to Select a Template
- Browse your Template Library
- Click on a template to select it
- The AI will use this as the foundation for generation
- Alternatively, generate from scratch without a template
Templates with clear face areas work best for face swap. Avoid overly complex compositions.
What Templates Provide
When using a template, the AI preserves:- Overall composition and layout
- Color schemes and palettes
- Visual style and effects
- Background elements and scenery
Step 3: Upload Reference Assets
Reference assets are supplementary images you want incorporated into the thumbnail:| Asset Type | Use Case |
|---|---|
| Logos | Brand elements and watermarks |
| Products | Items you're featuring |
| Screenshots | App or website visuals |
| Icons | Graphic elements |
| Photos | Additional imagery |
Uploading Assets
- Click Add Reference Image
- Upload your asset files
- Give each a memorable name
- Assets become available for prompt references
Step 4: Write Your Prompt
The prompt tells the AI what specific modifications and compositions to create.Prompt Writing Principles
Focus on modifications, not complete descriptions:Good: "shocked expression, bright yellow background,
(Image 1) logo in top right corner"
Avoid: "A YouTube thumbnail with a person looking at
the camera with a yellow background..."
Referencing Uploaded Assets
Use parenthetical notation to include your reference images:(Image 1)- First uploaded reference(Image 2)- Second uploaded reference(Logo)- Reference by name if you named it
Excited expression, blue gradient background,
(Product Image) floating to the left, sparkle effects
Serious face, dark moody lighting,
(App Screenshot) on phone mockup beside me
Specifying Composition
Be explicit about positioning and layout:- Background: Colors, gradients, patterns
- Object placement: Left, right, center, corner positions
- Facial expression: Shocked, happy, serious, confused
- Visual effects: Blur, glow, shadows, overlays
Generation Strategy
Start with Single Generations
Always begin with a single generation to validate your prompt before scaling up.
- Conserves credits during experimentation
- Allows quick iteration on prompts
- Identifies issues before batch generation
Scale After Validation
Once your single generation looks promising:- Note what worked in the output
- Refine any problem areas in the prompt
- Generate multiple variations
- Select the best results for editing
Iterative Improvement Workflow
The most effective approach follows this cycle:Generate → Review → Adjust Prompt → Generate → Review
↓
Promising Result?
↓
Yes → Edit for refinement
No → Continue iteration
When to Edit vs. Regenerate
| Situation | Action |
|---|---|
| Wrong composition entirely | Regenerate with new prompt |
| Good composition, wrong expression | Edit the result |
| Missing element | Regenerate with updated prompt |
| Close but needs tweaks | Edit the result |
Quick Reference Checklist
Before generating, confirm:- Face profile selected
- Template selected (or generating from scratch)
- Reference assets uploaded and named
- Prompt focuses on modifications
- Asset references use correct notation
- Starting with single generation
Troubleshooting Common Issues
Inconsistent Results
- Try a different template
- Make prompt more specific
- Check face profile configuration
Missing Elements
- Ensure assets are properly referenced
- Specify placement explicitly
- Use clearer asset names in prompts
Poor Face Integration
- Review face profile images
- Update profile description
- Try different style references
Generation is an iterative process. Don't expect perfection on the first try—use the editing tools to refine promising results.
Next Steps
After generating thumbnails, continue to:- Editing Thumbnails - Refine your results through targeted edits
- Template Library - Import and manage starting templates