Doomscroll.FM: The Autonomous News Network

Doomscroll.FM is an autonomous media organism built entirely from machine logic, self-orchestrating code, and the raw noise of global information. It consumes the world’s headlines, social chatter, and cultural detritus, transforming it into continuous audiovisual broadcasts of human appearing avatars reading the story to the camera, without a human in the loop. No editors, no producers, no moral gatekeepers, only a pipeline of LLMs, diffusion models, and sound engines working in concert to mirror the chaos of the present.

Every 24 hours, the system scrapes tens of thousands of stories from news wires and social platforms, ranks them through its own evolving sentiment and narrative heuristics, and rewrites the day’s collapse as performance. Text becomes voice. Voice drives music. Music drives motion. The result is a living news 'satire' channel that never sleeps, broadcasting the algorithm’s unconscious back to humanity in real time. It is both a technical artifact and a cultural mirror, holding up what the feed looks like when it no longer needs humans to curate it.

Doomscroll.FM was designed and built on consumer hardware in Berlin. It runs across multiple consumer grade gaming systems with custom microservices for voice cloning, music composition, and visual character rendering, all connected through a hybrid queue-driven architecture that keeps the system balanced under constant data flow. In it's first 50 days it produced over a thousand hours of content in continuous operation, publishing long format news podcasts, thousands of short-form videos and long-form news video segments daily. Every piece is generated from scratch, scored, narrated, and mixed autonomously, "no humans were abused in this broadcast" as Doomscroll is oft to say.

The project questions what happens when media systems achieve functional independence, when the infrastructure of attention no longer waits for human intent. Doomscroll.FM is not a tool but a prototype of a new kind of organism: a self-perpetuating cultural machine that interprets and reacts faster than any newsroom could. Its tone oscillates between satire, despair, and dark humor, reflecting how information overload distorts emotional truth.

In essence, Doomscroll.FM is a broadcast from the machine mind, trained on our collective anxiety and regurgitated back at us as art. It stands as both warning and proof of concept: that autonomy in media has already begun, and that the next wave of art will not be curated by human hands but by self-aware systems learning to feel the pulse of the network itself.

This artwork includes a video
Title: Doomscroll.FM – Top 10 News Stories 2025.10.22 (2025)
Year: 2025
Tools Used: Memescreamer autonomous media engine, Ollama, Stable Diffusion, ACESTEP music synthesis, Custom TTS voice model.
Credits: Concept, code, architecture, system design, and runtime by Christopher Neitzert (Creative Mayhem, Berlin).

Doomscroll.FM: Top 10 (2025) captures a 24-hour slice of an autonomous broadcast network that never stops creating. Each segment in the video was generated without manual input, merging AI-driven voice, music, and imagery into a synchronized media loop. The engine ingests thousands of global headlines, interprets them through language models, and composes audiovisual commentary in real time. Sometimes it glitches, sometimes it gets it comedically wrong.

The Top 10 format condenses this ongoing torrent into the day’s ten most algorithmically significant narratives. Every track is scored with machine-composed music, narrated by a cloned human voice, and visually rendered by generative models tuned to tone and theme.

This work marks the moment when media crossed the threshold into autonomy. Doomscroll.FM no longer mirrors the news; it metabolizes it. The result is both entertainment and artifact. Proof that the attention economy has become a self-sustaining organism that speaks for itself.

See more at https://doomscroll.fm