Learn how to create podcast-style thumbnails featuring two people facing each other
Podcast thumbnails typically feature two people facing each other with text in the center. This guide covers the techniques for generating and iterating on this specific layout.
Start with an inspiration thumbnail that demonstrates your desired podcast layout:
When writing your prompt, always use the exact person name from your person library. The AI matches the name you write to the person you selected.
Example: "Replace the right person with Alex"
Be explicit about which side each person should appear on. Use clear directional language in your prompt.
Example: "Place Sarah on the left and Mike on the right. They look at each other."
Request varied expressions to avoid repetitive-looking thumbnails across episodes:
Different expressions help each episode feel unique while maintaining a consistent layout.
You can modify the text while preserving the visual style:
Example: "Change the text. First line: 'The Future of AI'. Second line: 'Episode 42'. Keep the exact same style."
When replacing both faces in a reference, use the pixelate technique for cleaner results:
Example: "Replace the person on the left with Sarah. The person on the right is Mike. The two people look at each other. Same style shadow and contrast."
Always distinguish left from right explicitly. Vague references like "the other person" lead to unpredictable results.
Double-check that the names in your prompt match exactly with your person library entries. The AI uses name matching to identify which person to render.
For natural-looking podcast thumbnails, specify how the people relate to each other:
When swapping both people in a reference, pixelating the original faces forces the AI to generate completely new faces instead of trying to blend with existing features.
