Midnight Stranger: Your Local spooky story teller

Description and use

Midnight Stranger is an offline horror radio that generates spooky stories with a local Gemma model, narrates them via Piper TTS, and mixes synthetic ambience, all served through FastAPI and a browser UI.

Problem

Provide an offline horror radio that generates and narrates spooky stories without internet or cloud dependencies.

Outcome

Midnight Stranger provides a fully offline horror radio experience, generating stories with Gemma 3, narrating with Piper TTS, and mixing synthetic ambience via numpy, all served through FastAPI and displayed in a single‑page web UI.

Workflow

• Open the web page and press "Turn the radio on".
• Select a station via the browser dial.
• The FastAPI backend queues a request to the local Gemma 3 model to write a story.
• The story is passed to Piper TTS to generate audio.
• Numpy generates synthetic ambience and mixes it with the narration.
• Captions are generated from the story text and displayed.
• The audio stream is played back in the browser.
• The process repeats for each station change.
• All components run locally with no internet required.

Evidence

Midnight Stranger is a horror radio that works with the internet switched off.

Source

Nothing on that dial is a recording. A local language model wrote each story a few minutes earlier. A local voice model read it aloud.

Source

Built for the Hacktoberfest Open-Source AI Challenge, Week 4: Ghost in the Machine.

Source

Works with the lights (and the internet) off. Everything runs on your own machine.

Source