Editorial workflow

Private proof of concept

Reporter

A local review workspace that turns long-form recorded speech into a structured, editable transcript without hiding what the automation changed.

The problem

Why it exists

Transcription is a long and labour intensive process. Reporter seeks to take the sting out of this by taking a parliamentary recording from raw audio to a publishable record. It uses ASR to produce the first transcript, then AI to clean it up and perform a first edit before speaker names are added. Human editors take over from there: sittings can be divided between colleagues, edited in the same suite and peer-reviewed before publication. The audio stays aligned with the transcript throughout, with foot-pedal control for reporters who work that way. When the Official Report is finished, it can be published as JSON, XML, Markdown or DOCX.

What is implemented

The useful bits

  • 01

    Create local runs from recorded audio and track each processing stage

  • 02

    Transform provider output into structured speaker and timestamp segments

  • 03

    Map names and speakers, with reusable voice-profile tools

  • 04

    Compare automated edits, listen against source audio, and save paired review versions

  • 05

    Recover partial batch work and export versioned text and DOCX outputs

Current limits / no hand-waving

What this page is not claiming

Currently a private prototype and not yet available for public download.