How a deepfake is made and caught
Process map by Brian. 17 steps, 2 layers. Tags: ai, tech, misinformation, controversy.
- The maker: Collect photos, video and audio of a target
- The AI model: Train a model on their face or voice
- The AI model: Generate fake video or audio
- The maker: Polish: lighting, lip sync, compression
- The maker: Post it
- A platform: Does a label or a detector flag it?
- A platform: Label, limit or remove it
- A viewer: Does the viewer check the source?
- Detection tools: Look for provenance data or telltale glitches
- A platform: Fact-checkers and the person respond
- A viewer: Believes it and shares it
- A viewer: It spreads