What is Attention Mechanism?
The core component of transformer AI models that allows them to focus on relevant parts of the input when generating each part of the output.
How it works
Attention mechanisms allow AI models to dynamically focus on the most relevant parts of the input when generating each element of the output. When generating a video frame, the model attends to the relevant parts of your text prompt, the relevant parts of any reference image, and the relevant parts of previously generated frames. This is what allows the model to maintain consistency and follow instructions. Self-attention (attending to different parts of the same input) enables spatial coherence within a frame. Cross-attention (attending between text and image) enables prompt following. Temporal attention (attending across frames) enables temporal consistency in video. The quality of attention mechanisms is a primary differentiator between models.
Tools that use attention mechanism
Related terms
Frequently asked questions
What does attention mechanism mean in AI video?▾
Need AI video produced by professionals, not generated by yourself?
Apostle is an AI-native video production studio. We use every tool on this page in real client work.
Get in touch