TrainScription

About

TrainScription is a local AI transcription tool delivered as a Chrome extension. It captures microphone audio alongside browser tab audio or full desktop system audio, then transcribes speech in real time using Whisper running directly on the user's device through WebAssembly (transformers.js). Audio processing happens entirely in-browser, in volatile RAM-only chunks — nothing is sent to cloud transcription services, and no bot joins your call as a participant. The product emphasizes privacy and offline operation after the one-time model download, with full transparency about its only external connections (model download and license verification — neither touching audio or transcript content). A key differentiator is the Phonetic Brain: a user-trained vocabulary system. Record a quick sample of any name, acronym, product term, or case name, and the system captures the phonetic variants Whisper tends to mishear — then auto-corrects them to the canonical spelling on every future transcript, before you see it. The trained Brain is portable as a .train.json file. Transcripts are speaker-labeled (ME/THEM, customizable), timestamped, stored locally, and exportable as TXT or SRT. Recovery sessions are saved locally in case of crash or accidental close. Full Desktop Mode extends capture beyond the browser tab to any app making sound on the machine — Teams desktop, Zoom desktop, Discord, Slack calls, media players, and more. Not just browser tabs. Free tier: 15-minute sessions, 3 Brain slots, 3 Recovery sessions, per-segment export. Pro: $9.99 one-time, no subscription — unlimited sessions, Brain slots, and Recovery storage, one-click full export, cross-device access, all future updates included. Built for anyone who talks to people with unusual names or specialized jargon — researchers, lawyers, healthcare workers, GMs, journalists, remote teams — who need accurate, private, local-only transcripts without sending a word to the cloud.

Recognition this tool has earned on NextGen Tools.

• #3 of the weekTop launch on the weekly leaderboardJun 22, 2026

Problem & solution

Who it's for

Professionals, researchers, interviewers, and frequent meeting participants.

• Why Software Should Remember What You've Already Taught It

  Most software treats every interaction as though it's the first one. You correct the same mistakes, explain the same terminology, and repeatedly teach systems things you've already taught them before. But people don't work that way. Learning is cumulative. This article explores why software often forgets what users already know, why that creates unnecessary friction, and how TrainScription approaches transcription as an experience that improves over time instead of constantly starting over.

  NxGn Tools·Aug 3, 2026

• The False Divide Between Live Conversations And Recorded Ones

  Meeting transcription has traditionally focused on live conversations. But some of the most valuable knowledge we encounter isn't live at all. It comes from webinars, conference talks, interviews, lectures, podcasts, and recordings we've collected over time. This article explores why software often treats these as separate categories, why people rarely do, and how TrainScription approaches them as different paths to the same destination: understanding.

  NxGn Tools·Jul 11, 2026

• The Strange Way Digital Information Never Dies

  Physical information naturally disappears. Paper gets thrown away. Whiteboards are erased. Notebooks fill up and are replaced. Digital information behaves very differently. It accumulates almost effortlessly, often without intention or purpose. This article explores why "keeping everything forever" isn't always preservation, why forgetting can be valuable, and how TrainScription approaches the lifecycle of conversations differently.

  NxGn Tools·Jul 9, 2026

Share your thoughts about this tool.

No comments yet. Be the first to leave one.