How to Get The Best Results

The best way to generate a realistic deepfake is to provide the AI with very clean and diverse data, in the form of videos and images.

Some tips to improve the quality of your deepfake include:

• Ensure there is only one face in each frame of the video. If the AI detects multiple faces, the result will suffer tremendously. Please note the AI often picks up the back of someone's head as a face.
• Ensure the videos and images have multiple angles of the face, multiple expressions and a variety of lighting conditions. The more abundant and diverse, the better the result.
• Ensure both videos you are swapping are the same resolution. Our system works best when both videos are 1080p or better.
• More iterations: 10,000 iterations is the bare minimum required to get an adequate result, however, the truly impressive results come with 50,000 or more iterations. Training the same model with new data can make a huge difference.