Face2FaceAI — Quick Start Guide (Premium)

Face2FaceAI copies a facial expression from one photo (the Source) onto a face from another photo (the Target), then places the edited face (the Result) back into its original photo. You move left to right in the app through three bottom tabs: Input → Blend → Output. All processing happens on your device — photos never leave your phone. Tip: first time through, long-press any control for an instant explanation, or look up controls by name at any time in the Help Glossary in Settings.

Tab 1 — Input

  1. Load your Source — the face whose expression you want to copy. Tap the upper panel’s upload button → Gallery.
  2. Load your Target — the face whose expression you want to change. Tap the lower panel’s upload button → Gallery (or Camera).
  3. Each face is detected automatically (SingleFace Mode), intended for when there is one face per image. Other, more complicated input modes are available — for example, MultiFace Select Mode to select one Source or Target face among several in a photo, or Manual Location Mode to manually reposition a face detection BBox (Bounding Box).
  4. You can usually leave Mirror unchecked — use it when the Source and Target faces look in opposite directions.

Tab 2 — Blend

The three previews show your Source Face, Target Face, and Result Face. Set the sliders, then tap Run Blend (a few seconds to process). Before you run a blend with non-zero settings, Result Face just shows a copy of Target Face.

  1. A recommended starting point for blend sliders is: Eye Expression 0.25, Other Features 0.75, Head Position 0.0, Feather Radius 10.
  2. If facial features, especially eyes, look distorted, switch to: Eye Expression 0.0, Other Features 0.5, Head Position 0.0, Feather Radius 10.
  3. If the change in Result Face is too small, instead try: Eye Expression 0.5, Other Features 1.0, Head Position 0.0, Feather Radius 10.

Re-run as often as you like, adjusting in small steps for best results. For additional suggestions about possible fixes, check out Troubleshooting, also in Settings.

Tab 3 — Output

  1. Select Reinsertion Mode — this places the edited face back into the original Target photo for the most natural-looking result.
  2. Tap Finalize.
  3. If hair or background looks blurry around the face, or there are hard edges where the face is reinserted, go back and increase Feather Radius (Tab 2) by 5. Also increase Target BBox Size (Tab 1) by 0.1 or 0.2 if you see a double-exposure artifact. Then re-blend and finalize again.
  4. Save or Share when you’re happy with it.

Two Special Workflows

Using a Template as your Source. Instead of a gallery photo, tap the Source upload button → Expression Templates for a built-in library of expressions across five characters. For the closest match, pick the character whose face most resembles your Target. If one character’s version doesn’t blend well, try the same expression on a different character (“By Expression Name” in the menu). Templates often show a long neck — if your Target doesn’t, switch the Source to Manual Location Mode and move its BBox up so the neck is hidden; this noticeably improves results.

Fixing an asymmetric selfie (one face). Load your selfie as the Target. Then load the same selfie as the Source and turn on Mirror in the Source panel — the mirrored copy balances the two sides of the face. In Tab 2, run the blend with intermediate settings (around 0.5) for Eye Expression, Other Features, and Head Position, and Finalize with Reinsertion. If the eyes end up looking the wrong way, set Eye Expression to 0.0. If the corrected face looks stiff or blank, lower all sliders until it looks natural again — even a gentle correction visibly reduces asymmetry.