TranscriptX documentation¶
TranscriptX is a local-first workbench for people who want to think with transcripts. You import files you already have, run analysis on your machine, and keep the results.
It does not transcribe audio in the app. Bring JSON, SRT, VTT, or similar files, then use Import Transcript.
See how it works: first analysis — import the sample, run Balanced, read Overview.
Everyday jobs: Using TranscriptX.
Is this for me? How TranscriptX compares.
Privacy: files stay on your computer; optional local AI is Ollama and off by default.
The GitHub README is the same first-run story.
Start here
Workflows
- Identify and name speakers
- Investigate a question and trace it back to evidence
- Use local AI for synthesis
- Export a finished analysis
- Explore Charts for a finished run
- Bundle transcripts into a group
- Correct wording while reading
- Rename a managed transcript
- Browse longitudinal speaker profiles
- Assist naming across transcripts with voice
Using TranscriptX
- Settings, profiles, and analysis knobs
- Docker
- Directory watcher
- Auto-identify speakers
- Exporting runs (ZIP, HTML, EPUB)
- Workspace backup and restore
- Known limitations (1.0 programme)
- Analysis models
- Local LLM integration (Ollama)
- Corrections in the Transcript viewer
- Karaoke playback (Transcript viewer)
- Corrections Studio LLM discovery
- TranscriptX terminology index (non-authoritative)
- WhisperX standalone (optional reference recipe)
- Whisper-WebUI (optional interoperability recipe)
Advanced
Developers
- Developer documentation index
- TranscriptX Roadmap
- TranscriptX Architecture
- Developer Quick Start — TranscriptX
- Packages used
- Code and architecture reviews
- Contract boundary map
- Public surfaces contract
- TranscriptX Storage Policy
- Module Catalog
- Web launcher and Python API
- Lexical diversity analysis
- Keyphrases (
keyphrases) - Epistemic markers (
epistemic_markers) - Politeness markers (
politeness) - Topic-shift (
topic_shift) contracts — Wave 1 - ASR confidence (
transcript_quality)