Training Face Models

Face models let the AI learn what you look like so it can generate thumbnails featuring your face. Once trained, you can swap your face into any thumbnail with a single click.

Why Train a Face Model?

• Consistent branding — every thumbnail features the real you
• Save time — no need to upload a photo every time
• Better quality — trained models produce more accurate faces than one-off uploads
• Expression control — the AI can generate different expressions (shocked, happy, curious) while keeping your likeness

How to Train a Model

1. Go to Faces in the left sidebar
2. Click Train New Model
3. Upload 5 or more clear photos of your face
4. Give the model a name (e.g., "My Face" or your name)
5. Click Start Training

Training takes approximately 15 minutes. You'll be notified when it's ready.

Tips for Good Training Photos

The quality of your training photos directly affects the quality of your generated faces.

Do

• Use clear, well-lit photos where your face is fully visible
• Include photos from different angles — straight on, slight left, slight right
• Show different expressions — neutral, smiling, surprised
• Use photos with simple backgrounds that don't distract
• Include at least one close-up portrait

Don't

• Use blurry or low-resolution photos
• Use heavily filtered or edited photos
• Use photos where your face is partially hidden (sunglasses, masks)
• Use group photos where the AI can't tell which face is yours
• Use photos that are all from the same angle

Managing Your Models

• You can train multiple face models (one per person)
• View training status on the Faces page — models show as "Training" or "Ready"
• Each model can be used in the editor's Face Swap tab or during AI generation
• Delete models you no longer need from the Faces management page

Using Your Trained Model

Once trained, your face model appears in:

• The Face tab in the thumbnail editor — for swapping into existing images
• The AI generator — for face-aware generation from scratch
• The Channel Makeover tool — for batch thumbnail regeneration