Diffusion model
The kind of model behind most image and video generators: it starts from noise and removes it step by step until a picture the prompt describes is left.
It is trained by the reverse of what it does. Noise is added to real images until nothing is left, and the model learns to undo one step of that; generation is that learned step run many times over from pure noise.
Two consequences follow for anyone paying the bill. The number of steps is a dial between speed and detail, and video costs far more than the frame count suggests, because consistency between frames is work the model has to do on top of drawing each one.
It is a different family from the text models. Prompts do not carry across: a diffusion model has no conversation, no tools, and no memory of the previous image unless it is handed one.