Redactor v7.3: speaker-level audio redaction
Redactor 7.3 introduces speaker diarization, so reviewers can redact one person's voice across a recording instead of muting the whole track. It ships with a rebuilt speech engine, a redesigned transcript editor, and a stronger audit archive.
What's new
Redact one speaker, not the whole track
Redactor can now attempt to separate who is speaking in a recording. Speakers are identified automatically, and reviewers can label, correct and merge them — then redact a specific person while leaving everyone else audible.
This is the difference between releasing an interview recording and shelving it. A bodycam clip with an officer, a complainant and a bystander used to mean either muting long stretches or working through the audio by hand. Now the reviewer works at the speaker level.
Speaker diarization ships as beta. Results vary with recording conditions — overlapping voices, background noise and mic placement all affect how cleanly speakers separate. Every result should be reviewed before export. We're refining it across the next several releases.
A rebuilt speech engine
Speech transcription runs on a new engine in 7.3, with a redesigned editor around it: a maximised view alongside the existing sidebar, search across words, regular expressions and speakers, selection inversion, transcript export in raw text and JSON, and live word highlighting during playback.
Installers and Docker images are smaller than in previous releases. English ships in the box; additional languages download after installation rather than shipping with it.
Work time tracking
Redactor now records how long an editor actively spends on each session, with a start/stop clock and a live readout in the editor, plus a running total in the server video list. Supervisors can reset a video's recorded time, and the data exports with the session's audit logs.
For teams billing redaction work or reporting review effort to an oversight body, this replaces a spreadsheet.
Also in this release
The audit archive now includes a manifest recording the archive's origin — host, network addresses and user — plus a SHA-256 checksum of each session's audit log alongside the video's metadata
The speech editor stays responsive on hour-long recordings with thousands of words
Waveform zoom redraws smoothly during rapid mouse-wheel and pinch gestures
Playback-speed controls behave consistently across browsers and the desktop app
Manual Live Track recordings survive a server disconnect — in-progress recordings are staged and committed automatically (server deployments; desktop was unaffected)
Sessions are preserved when video files go missing, and reconnect automatically if the files return
Licence seat availability now calculates correctly for licences with a fixed expiry date
Availability
Redactor 7.3.0 is available now for desktop and server deployments. Existing customers can upgrade from the download page; your sessions and redaction data carry forward.
Full technical detail is in the 7.3.0 release notes.
See speaker diarization in a live session.
