Requires persona training first. This feature generates images using your trained AI likeness. If you haven't uploaded photos and trained your persona yet, start there →
How prompts work
You describe the vibe, we generate options.
Good prompts
- "Surprised expression, looking at phone, clean background"Clear, specific, achievable
- "Professional headshot, slight smile, navy blue background"Simple pose, clear context
- "Excited, pointing at something off-screen, bright lighting"Expression + gesture + lighting
Prompts that struggle
- "Me fighting a dragon on a mountain"Too complex, face will drift
- "Close-up of hands typing on laptop"AI hands are still unreliable
- "Me with 5 other people in a meeting"Multiple faces = chaos
Beyond the face
Thumbnails aren't just about the photo. After generation, you can:
- • Add text overlays — headlines in your brand fonts, positioned for readability
- • Apply brand colors — automatic if you've set them up, or pick custom
- • Swap backgrounds — solid, gradient, or generate a contextual one
- • Add visual elements — arrows, circles, emoji, basic graphics
It's not Photoshop. But for "scroll-stopping thumbnail in 2 minutes," it's often enough.
The honest truth about AI faces
AI face generation is good, not perfect. Expect to see:
- • Your face looking slightly "off" in some generations
- • Weird hands if they're in frame (AI's achilles heel)
- • Inconsistency between generations
- • Occasional artifacts in complex scenes
The workflow is: generate 4-5 options, pick 1-2 that work, discard the rest. For most creators, getting 2 usable thumbnails in 3 minutes beats spending an hour in Photoshop or scheduling a photoshoot.
If every thumbnail needs to be perfect, you still need a designer. If "good enough, fast" is acceptable, this works.
The real benefit: consistency
When every thumbnail uses the same trained persona and brand colors, your content starts looking cohesive. Someone scrolling recognizes your style before reading the title. That recognition compounds over time. It's hard to maintain manually — you'd need the same photos, same editing approach, same colors every time. Automated, it just happens.