TrainScription is a Chrome extension that transcribes dual-channel audio (microphone plus browser or desktop audio) entirely on-device using Whisper AI via WebAssembly, with no cloud upload and no bot joining the call.
What is TrainScription?
TrainScription is a local-first transcription Chrome extension that captures two independent audio streams — your microphone and meeting audio from a browser tab or any desktop app — and transcribes them in real time using OpenAI's Whisper model running in WebAssembly. The entire process occurs on your device, with no network requests during transcription and no account required for the free version. Developed by Terrance Henry, it aims to keep sensitive conversations private while providing speaker-labeled, timestamped transcripts that improve over time via a built-in vocabulary trainer called the Phonetic Brain.
Key Features
- Phonetic Brain — A local correction layer that learns from your corrections: highlight a misheard word, assign the correct spelling, and every future transcript auto-corrects before you see it. Pre-train vocabulary on your phone using the free BrainTrainer companion app.
- Browser Tab & Full Desktop Capture — Capture audio from any browser tab (Google Meet, Zoom web, Teams web) or from any desktop app (Teams, Zoom, Discord, etc.) via system audio sharing.
- Dual-Channel Audio — Your microphone and meeting audio are captured as two independent streams, preserving overlapping speech and labeling each channel with configurable names.
- On-Device AI — Whisper runs locally via WebAssembly after a one-time model download; no internet connection needed for transcription, no API key, no server.
- No Bot Joins the Call — The extension captures audio from your machine's audio pipeline, not as a meeting participant.
- Session Recovery & Export — Sessions are automatically segmented and stored locally; export as .txt or .srt files, or use Pro's one-click assembled session export.
- Pause Controls & Silence Auto-Stop — Pause individual channels or everything without ending the session; optionally auto-stop after a configurable silent period.
- Audio & Video File Import (Pro) — Process an existing recording through the same on-device pipeline, no upload required.
Who is it for?
- Legal, finance, and consulting professionals who handle sensitive conversations and need to avoid routing audio through third-party cloud services.
- Researchers and interviewers capturing browser-based interviews without third-party transcription services, keeping data local and exportable.
- Heavy meeting users who want to avoid monthly subscription fees — one $9.99 payment covers unlimited sessions.
- Users with specialized vocabulary (e.g., product names, medication names, case names) who need a transcription tool that can be taught to correct misheard terms automatically.
What can you do with TrainScription?
- Transcribe live meetings without a bot joining the call: open your meeting in a browser tab or desktop app, click Begin Capture, and watch a real-time speaker-labeled transcript appear.
- Correct misheard jargon on the fly: highlight any word the AI got wrong, assign the correct spelling, and the Phonetic Brain permanently fixes it for all future transcripts.
- Revisit past sessions: use Session Recovery to review, rename, and export any past segment or full session as .txt or .srt.
- Import existing recordings: with Pro, process an audio or video file you already have through the same local Whisper pipeline, applying your Brain corrections.
How does TrainScription work?
- Open your meeting in a browser tab or desktop app. The extension captures audio from your machine's audio pipeline — not from the meeting's participant list.
- Click Begin Capture and choose Browser Tab mode (to capture a tab's audio) or Full Desktop mode (to capture any app making sound). The extension captures two audio streams: your microphone and the meeting audio.
- Audio is processed in 5-second volatile chunks through a local Web Audio graph with silence gating. Whisper runs in WebAssembly on your device, with no network calls.
- Every 5 seconds, a new chunk is transcribed and appended as speaker-labeled, timestamped text. You can close the popup; transcription continues.
- If you see a misheard word, highlight it to open a popover and assign the correct spelling — the Phonetic Brain updates instantly and applies to future transcripts automatically.
Pricing
TrainScription offers a Free tier with a 15-minute session cap, 3 Recovery sessions, and 3 Phonetic Brain slots. Pro costs $9.99 once and unlocks unlimited session length, unlimited Recovery sessions, unlimited Brain slots, one-click assembled session export, Audio & Video File Import, Backup & Restore, and cross-device use via email login. No subscription, no per-minute billing.
Pros and cons
- Pros: Fully local and private (no cloud, no bot); one-time payment for Pro; Phonetic Brain permanently fixes jargon; works with both browser tabs and desktop apps; supports offline use after model download; dual-channel with speaker labels.
- Cons: Free tier has session caps; transcription may be slower on older hardware due to local inference; currently only available as a Chrome extension.
Alternatives
- Otter.ai — Cloud-based with a bot that joins meetings and monthly subscription ($16.99/mo).
- Fireflies.ai — Cloud transcription with meeting bot and per-seat pricing ($10/mo).
- Rev — Cloud-based with human or AI transcription, monthly fee ($9.99/mo).
FAQ
Is TrainScription free?
Yes, there is a free tier with no account required. It limits sessions to 15 minutes, stores up to 3 Recovery sessions, and allows 3 Phonetic Brain slots.
What does the Phonetic Brain do?
It corrects words that Whisper consistently mishears. Train it once by highlighting a misheard word and assigning the correct spelling. Every future transcript corrects that word automatically before you see the output.
Does TrainScription work offline?
Yes, after the initial one-time model download from the model provider, all transcription runs locally with no internet connection required.
Can I transcribe an existing recording?
Yes, Pro users can import audio or video files and process them through the same on-device pipeline — no upload needed.
Does a bot join the call?
No. TrainScription captures audio from your machine's audio pipeline, not as a meeting participant. No third-party participant joins the call.
