← All workSoftware · Offline desktop tool

Call Review

Turns a folder of recorded jail calls into a searchable report an investigator can open with a double-click.

Version 0.1, in private testingStatus as of November 12, 2026

Call Review is not publicly released and there is no download. The Mac build has been tested. The Windows build is out with a tester, and the Linux build is packaged but has not been run on a real machine yet.

  • Call Review call view with a two-channel waveform, transcript, and notes panel, showing demo data
  • Call Review timeline with call counts, talk time, and a calls-per-day chart, showing demo data
  • Call Review highlights view listing watched keywords and how often each is mentioned, showing demo data

Every screenshot is from a demo report built on synthetic data. The names, numbers, and conversations are invented, and no real case material appears on this page.

What it does

Built for a computer with no internet

A jail-call export can be hundreds of recordings and a spreadsheet. Call Review, part of a set we call Detective Tools, reads an export from the Securus phone system, transcribes every call on the local machine, and builds a report where an investigator can search, listen, bookmark, and take notes.

It is built around one rule: nothing connects to the internet. Transcription runs on the workstation. The finished report is plain HTML that opens from a folder on a locked-down computer, with no server, no install, and no network.

  • Offline transcription with each line labelled inmate or other party, and the recorded announcement skipped
  • A player with a two-channel waveform that follows along with the transcript
  • Phone numbers, money, dates, names, and addresses picked out of every transcript
  • A starter list of 413 keywords in 11 groups, plus your own keywords for the case
  • Contacts, timeline, search, highlights, and evidence views across the whole export
  • Calls to numbers marked attorney or privileged are hidden by default
  • Notes and bookmarks that save to a file and load back in
  • MP3 clip export with a SHA-256 list for the evidence record
  • Optional shading that shows where the transcription model was unsure

At a glance

Status
Version 0.1, in private testing
Runs on
An offline workstation: macOS today, Windows in testing, Linux packaged
Network
None. Connections are blocked in code and checked by an automated self-test
Testing
More than 830 automated tests

Built with

  • Python
  • Local speech-to-text (faster-whisper, large-v3-turbo)
  • Optional NVIDIA GPU
  • Plain HTML, JavaScript, and CSS report
More software

Other things we have built

Live on the App Store and Google Play

Book Spotter

iPhone and Android app that uses the camera and augmented reality to spot the books you are looking for on any shelf, offline, entirely on your device.

Details →
Submitted to the App Store, in review

Nertz Score

iPhone score keeper for the card game Nertz (also called Nerts, Pounce, or Racing Demon). Works offline, with no accounts and no server.

Details →
In development, not released

ThriftSpotter

iPhone app for resellers: photograph a thrift-store shelf or an estate-sale table and it flags anything likely to resell above your minimum price.

Details →
In daily use here, public release planned

Tally Hour

Time tracker for Mac, iPhone, and the browser, synced through one small server. We run the business on it every day.

Details →

Need software like this built?

We scope it, build it, test it properly, and keep it running. You deal with the engineer who writes the code.