Using TranscriptX

Short, outcome-focused walkthroughs for real jobs in the web UI. They complement the reference guides; they do not replace them.

Use the same sample transcript across the set so the story stays continuous: planning_review.json (synthetic three-speaker launch planning meeting).

The still images are a Balanced run of Rapid Response Webinar Analysing the Biodiversity Beyond National Jurisdiction Treaty, after display names were assigned in Speaker Identification. Follow the steps with the sample file if you do not have that webinar.

Common workflows

Start here. First-time users should follow 1 → 5 in order.

#

Workflow

Outcome

1

First analysis

Import a transcript, name speakers, run Balanced analysis, and read Overview

2

Identify and name speakers

Turn diarized labels into readable names before using speaker-level results (optional Apply auto-identify; USB: inbox-watch --auto-name)

3

Investigate with evidence

Answer a concrete question and trace it back to the transcript

4

Local AI synthesis

Use optional local AI for summary and meeting extracts

5

Export results

Package a finished run as a ZIP export with HTML (and EPUB when available)

Workflows 1–3 and 5 do not require local AI. Workflow 4 does.

More workflows

Jump in when you need these tasks.

#

Workflow

Outcome

6

Explore Charts

Open run-scoped Charts for visual module outputs

7

Bundle into a group

Analyse several transcripts together

8

Correct while reading

Propose word/span fixes in Transcript Correct mode

9

Rename a transcript

Give a library transcript a clearer file name

10

Browse speaker profiles

Open speaker profiles linked across transcripts

11

Assist naming with voice

Enrol all + Pre-load so later transcripts propose people to confirm

12

Archive a transcript

Hide it from pickers without deleting; find it again from Library or Search

Prerequisites

  • TranscriptX web UI installed and running (Installation or Docker).

  • For workflow 4 only: local Ollama configured (LLM).

  • For workflow 10: at least one longitudinal profile (create via Speaker Identification), or follow the empty-state path in that guide.

  • For workflow 11: speaker_match extra, voice privacy enabled, seed named+linked profiles, and audio on managed transcripts.