Irreplaceable Biography

Artificial intelligence systems used in filmmaking produce media by their ability to resemble a dataset. Deep generative models are nowadays used pervasively, and language has become a mechanism to leverage production. Extending on recent advances of image diffusion, a model’s ability to reconstruct reality is here examined as a signature of aesthetic potential, where abstraction and bias is granted over an input.

Imposing a descriptive ground, Walt Whitman’s A Song of Myself (1892 version) was used to guide a 4-minute 20-second sequence, where imagery is developed from a narration. Orchestrating a cooperative and multimodal simulation, we probe the role of vision produced artificially with both experimental and narrative potential, followed by internal monologue. As generative modelling progresses new simulations are useful compositionally, and the same way a composer is of a certain context and time, a model is of a certain dataset.