Forge: Train Custom AI Models

Train a personal AI model from your own photos. Generate consistent characters, styles, and branded visuals from a model that’s yours.

What You Can Do

• Consistent Characters: Keep the same identity across many images and scenes.
• Brand & Style: Generate material that matches your brand’s look and feel.
• Product Shots: Create varied angles and environments without new photoshoots.

How It Works

1. Upload 10–20 images with varied angles
2. Start training and wait a few minutes
3. Open Create and select your new model
4. Generate images and videos with prompts

Prepare a Strong Training Set

A useful custom model starts with clear, varied source images. Choose photos that show the subject from different angles, distances, expressions, and lighting conditions while keeping the identity or visual style consistent. Avoid near-duplicate frames, heavy filters, large watermarks, and images where the subject is hidden. A focused set helps the model learn the qualities you care about instead of memorizing one pose or background.

After training, begin with a straightforward prompt and compare several outputs before adding complex styling. Keep the subject description stable while changing location, wardrobe, camera language, or mood one element at a time. This makes it easier to see what the model learned and to develop a reusable prompt pattern for portraits, product scenes, characters, campaign assets, and other consistent creative work.

From Training to Creation

Your trained model appears in Create alongside Fiddl.art’s base models. Use it for new images, test multiple aspect ratios, and keep the strongest results as references for later ideas. Only upload material you have permission to use, review each result before publishing, and keep private work private when it should not appear on a public creator profile.