Foundation Feature

Train AI on your face

Upload 10-20 photos. We fine-tune an image model on your appearance. Then you can generate new images of "you" for thumbnails, graphics, whatever — without scheduling another photoshoot.

The process, step by step

1

Upload 10-20 photos of your face

Different angles, different expressions. Good lighting. Recent photos that look like how you want to appear. We'll tell you if any need replacing.

2

Training runs (~15-20 minutes)

Close the tab if you want. We email you when it's done. The model learns your facial features, how light hits your face, your typical expressions.

3

Generate images whenever you need

Describe what you want: "surprised face," "professional headshot," "excited pointing." Get multiple options. Pick what works. Use for thumbnails, graphics, social content.

What makes good training photos

Better input = better results. Quality matters more than quantity.

Include these

  • • 3-4 shots looking directly at camera
  • • 3-4 shots at 3/4 angle (slightly turned)
  • • 2-3 profile or near-profile views
  • • Mix of expressions (neutral, smiling, animated)
  • • Good lighting (natural light or well-lit room)
  • • High resolution (512×512 minimum, higher is better)

Avoid these

  • • Blurry or motion-blurred photos
  • • Heavy Instagram filters
  • • Sunglasses covering your eyes
  • • Group photos cropped down
  • • Old photos with different hair/look
  • • Extreme angles (looking straight up/down)

Setting expectations

AI-generated faces are impressive but imperfect. You will see:

  • • Generations that look "almost you" but slightly off
  • • Hands that look weird (AI's universal weakness)
  • • Inconsistency between different generations
  • • Some expressions that just don't work

The workflow is: generate several options, pick the ones that look right, ignore the rest. If you get 2-3 usable images out of every 5 generations, that's normal and still faster than alternatives.

If you need photorealistic perfection every time, you need a photographer. If you need "good enough for a thumbnail" fast, this works.

Your data, your control

We take this seriously because it's your face:

  • • Your photos and model are private to your account
  • • We don't use your data to train any shared models
  • • You can delete everything at any time
  • • Encrypted storage for all uploaded images

What this unlocks

Once your persona is trained, you can:

Generate thumbnails with your face
Create social graphics featuring you
Consistent visual identity across content
Quick iterations without new photos

Train your persona

10-20 photos. 20 minutes of training. Then generate images of yourself whenever you need them.