Whisper listens
Speech-to-text runs locally with Whisper. Audio does not need to leave the network for the transcript to exist.
Field note 03 / homelab build
Yell…orNot
Voice transcription that stays on your own hardware.
Self-hosted / private
Whisper in. Piper out. No public signup, no cloud upload, no subscription meter.
LOCAL AUDIO LOOP
voice→Whisper→text→Piper
YOUR MACHINE / YOUR AUDIO / YOUR RULES
Why it exists
Some words should not need a round trip.
Yell…orNot is a self-hosted voice layer for transcription and speech. It gives the audio work a private place to land, with the home lab doing the processing instead of a third-party API.
The shape of the system
Piper answers
Text-to-speech runs through Piper, giving the system a local voice instead of handing the response to a hosted speech API.
The meter stays off
No subscription, no per-minute cloud cap, and no vendor dashboard deciding how much of your own voice you can use.
Current boundary
Useful at home. Not packaged for strangers yet.
The current build runs as a private homelab service on CPU-only hardware. The product work ahead is packaging, documentation, and a safe interface for people who want local voice tools without becoming their own infrastructure team.