Our mission

Providing innovative technology for the preservation of radio heritage.

A Time Machine For Radio

Imagine a radio time machine that allows you to tune in to any AM or FM channel and listen live. Then staying on that channel, shift to any moment in the last decade by adjusting the year, month, day, hour, and minute on an additional time dial. Then add in the ability to perform text searches across all those years of broadcasts for any phrases in programs or advertisements and immediately access them for playback. (see demo below)

We have built a portable radio spectrum recorder that gives you this freedom. You'll have direct access to every moment of the broadcast past, including talk shows, news, commercials, and music. All of this is made available through text-searchable, time-encoded transcripts, accompanied by the original audio for analysis and research.

The Los Angeles / Long Beach archive

A living library of the airwaves

We are quickly approaching our first one million hours of continuous recordings with hundreds more being added each day. This is a truly unique competitive advertising resource.

The content will eventually be part of a digital museum for the general public and historical research. Contact us (below) for access or be the first to sponsor a new geographic area.

The Collector Technology

Lowering collection costs means more complete archives.

Our custom software runs on commodity pc hardware / linux and uses consumer grade SDRs. It is capable of capturing any radio spectrum from 100 Khz to 6 Ghz, equipped with real-time, software-based demodulation capabilities for AM, FM, HD radio, and SSB. It can perform these functions at a fraction of the cost of industry standard methods, which often require one tuner per channel or multi-tuner cards. We can collect the entire FM or AM band (as quadrature data) with a single software defined radio. All of this with zero commercial license dependencies or monthly fees. All processing takes place on the appliance so no internet access is required even when real-time speech-to-text transcription is configured. This enables extremely remote deployments. Contact us (below) to license your own collector.

The Transcriber Technology

Early in the radio-archive project we were faced with the high market costs of transcribing millions of hours of continuous audio. Just for the first 500,000 hours we had already recorded it would have cost over $253,000 using AWS Transcribe (based on Amazon's own cost estimator). Over 50 cents per hour and that's just for the transcribe calls; not even the whole batch processing solution! We needed to figure out how to do this and keep the total costs closer to 1 cent / hour yet still do it quickly. And if we could leverage any old Nvidia gaming GPUs we had lying around perhaps we could bring this cost down even further.

So we set upon building a self managing (whisper based) distributed transcription processing appliance strictly to solve our own problem. The whisper part is in ()'s because the transcriber part is swappable but we have not found a faster/more accurate option than (OpenAI's) Whisper . This appliance saved us over $200,000 in processing costs! It even scales up and down automatically as we add/remove GPU hardware from our pool. If you need to transcribe 10's of thousands to hundreds of thousands of hours per month with high accuracy (from 86% - 96% depending on settings), it could save you too. (contact us directly for details )

Run in your own datacenter

To use our techology for your own recordings or to make your own batch transcription jobs at a fraction of what you pay now, see the email contact information below or message us at https://www.linkedin.com/company/radio-archive-org

Development timeline

  • 2015: Began research into the full spectrum radio collector design.
  • 2019: (November) Software engineering begins.
  • 2020: Deployed our first FM prototype collectors in the Los Angeles and Orange County areas. These are still running today alongside later generations that are currently in testing.
  • 2021: Our first software defined AM radio collectors were completed and deployed.
  • 2022: Ongoing research doubles the recording channel density from our 2020 prototype version further decreasing costs of deployment.Also collector uptime now at 99.998% as technology matures.
  • 2022: A text-based search interface was incorporated to enable direct access to live and archived radio broadcasts by the deaf and hearing impaired. This feature allows all users to track discussions and themes across various channels and time.
  • 2023: Added support for realtime GPU based speech to text transcription coupled with the ability to do text, time, channel and locale based queries against them. Then deployed this as an appliance to our first commercial customers in the U.S. See the section on transcriber processing below for more details .
  • 2024: Refined our collector hardware replacing many of the analog components with custom realtime software making the system smaller, more durable, and more scalable.

Audio and Podcasts

Podcast format discussion of radio-archive.org (generated by NotebookLM)

Demos

Video demonstration: Searching the archive’s transcripts for specific phrases.

Older overview demo that includes the time based listening interface

Our collectors and transcribers can be licensed to run on-premises at your own site on your own hardware.

Contact us for access!

admin@radio-archive.org