Raspberry Pi Radio Transcriber
Budget: $30 – $250 USD
I need software that runs on a Raspberry Pi (or similar single-board computer) and continuously listens to incoming radio traffic fed through the 3.5 mm audio jack, then turns that speech into text. The stream is only in English and I want the final transcripts to be highly accurate—think Whisper-large, Vosk with custom language model, or any other cutting-edge ASR stack that you feel confident deploying on ARM hardware. A brief processing window is fine; transcripts can appear a few seconds behind the live audio, but anything noticeably longer will not meet my needs.
The Pi will be headless and unattended, so the solution should launch automatically on boot, recover if the audio source drops, and expose the running text through a simple method—local web page, REST endpoint, or even a continuously updating log file works as long as it can be consumed by another application.
Deliverables:
• Fully commented code or container image ready to flash or pull onto the Pi
• Step-by-step setup guide covering OS image, package installs, model downloads, and how to start/stop the service
• Test procedure I can follow to confirm accuracy with sample radio clips
If you’ve already optimized heavy models for low-power devices, let me know—performance per watt is a bonus.
The Pi will be headless and unattended, so the solution should launch automatically on boot, recover if the audio source drops, and expose the running text through a simple method—local web page, REST endpoint, or even a continuously updating log file works as long as it can be consumed by another application.
Deliverables:
• Fully commented code or container image ready to flash or pull onto the Pi
• Step-by-step setup guide covering OS image, package installs, model downloads, and how to start/stop the service
• Test procedure I can follow to confirm accuracy with sample radio clips
If you’ve already optimized heavy models for low-power devices, let me know—performance per watt is a bonus.