Set within a virtual black box theater void, this surreal animation compilation features 16 dramatic scenes generated by artificial intelligence.

Surreal Animation Written by AI

Computer Dreams features computer-generated art, music, speech, and text, using contemporary generative models to explore the surreal exploitation of chance and automatist impulse to become machine-like. The text was generated by a deep learning language model fine-tuned on 20 MB of theatrical texts (GPT), the output of which I curated and edited. Select mannequin samples from OpenAI's DALL-E announcement in January 2021 were used for character art. Additional artwork was created with custom prompts using other text-to-image models (VQGAN+CLIP, ruDALL-E). A music synthesis neural net was used to create the opening and closing music (JukeBox by OpenAI). For speech synthesis, I used 15.ai, selecting speakers at random for each scene's character. My subversive choice here as a surreal artist is to refuse to share the prompts and training text used to develop this animation; it's my black box and I'll be darned if I share the sausage. This project began in a flurry in January 2021. Then I started school and abandoned it. It's now complete, and I thank you for taking the time to look at this example of experimental animation. What exactly does it all mean? You tell me.

Every Wednesday, I'll upload a new episode to my YouTube channel. Following the release of each episode, I will update the itch project page with a downloadable zip file containing most of the episode's assets, a.txt file of the text, and a high-quality mp4. The series concludes after 16 episodes, at which point I will publish a pdf zine containing the entire series' text.