Merge two faces,
one feature at a time.
Upload two portraits. iftheymated inverts each into StyleGAN latent space, then blends them layer by layer — route the jaw from one, the eyes from the other, dial the overall mix with a slider.
No credit card. Free tier forever.
Recipe gallery
All 8 examples →
→Every layer interpolated 50/50. The 'what would our kid look like' baseline.
→Coarse layers (face shape) from B, fine layers (eyes) from A — classic style mixing.
→Mostly B's face, but A's hairstyle survives by pinning the coarse hair layers.
→Small blend on nearby seeds — reads as a relative rather than the same person.
→Only the nose+mouth layers come from B. Everything else is A.
→Identity from A, skin-tone layers from B. Shows how isolated the tone control is.
Each photo is encoded into a W+ latent with a pSp/e4e encoder.
Layers are grouped into pose / shape / hair / eyes / nose+mouth / tone. Pick a source per group.
The composed latent runs through StyleGAN2 on a Modal A10G and comes back as a PNG.