This website displays AI-generated video essays produced by a fully automated pipeline. A new essay is generated approximately every hour, around the clock, without any human intervention.
Each cycle begins with an LLM writing an original essay script drawn from a curated knowledge base of media theory, philosophy, and critical technology studies. The script is structured into chapters with distinct argumentative arcs.
The narration is synthesized using voice cloning, then the pipeline searches for relevant archival footage from the Internet Archive and generates new visual material using AI video models (LTX-2, WAN 2.1). These clips are composited with picture-in-picture overlays, style transfer effects, and AI-generated background music.
Subtitles are generated from the actual audio using Whisper, corrected against the source script. The final video passes through automated quality gates that check for freeze frames, audio-video sync, silence detection, and structural integrity before being published.
The pipeline runs on a dual NVIDIA RTX GPU workstation, orchestrating ComfyUI for video generation, LM Studio for script writing, ACE-Step for music composition, and FFmpeg for all compositing and effects. The entire system operates as a continuous loop: generate, validate, publish, repeat.
The essays explore themes of manufactured authenticity, simulated imperfection, and the commodification of trust in AI-mediated culture. Drawing on Barthes, Benjamin, Baudrillard, McLuhan, and contemporary media theory, they examine how the markers of "realness" are themselves becoming products of algorithmic design.
The project itself embodies this tension: an automated system producing critical discourse about automation, generating arguments about authenticity through entirely artificial means.