TranscribeGeek 1.1.2

Fixes transcription.

TranscribeGeek could install, download a model, and then fail on the first file with "Native Library not found in default paths". Every released version until now had this: 1.0.0, 1.0.1, 1.1.0 and 1.1.1. If you have been trying to make it work, it was not you.

What was wrong

The app ships as one executable with the speech library packed inside it. Whisper.net does not load that library by name; it goes looking for a runtimes folder, and every place it looks is worked out from where the executable sits. There is no folder next to a single executable, so it found nothing.

1.1.1 tried to fix that by finding the folder Windows unpacks the bundle into and pointing the library loader at it. That was still a bet on the host putting the files where the code expected, and on Windows it did not pay off.

What is different now

The speech library is carried inside the application itself. If it cannot be found beside the app or where the host unpacked it, TranscribeGeek writes its own copy into your profile and points the loader straight at it. Under two megabytes, written once, checked by size and SHA-256 afterwards so an interrupted first run cannot leave a broken file behind.

That route does not depend on the host doing anything, and it now runs on every build with the other two routes switched off, so it is exercised on the build machine rather than on your computer.

If it ever does fail again, the message now says where the app looked and how it got there, instead of only saying that something was not found.

Downloads below: TranscribeGeekSetup.exe (installer) and TranscribeGeek-portable.exe (no install needed). SHA256SUMS.txt lists the hashes for both.

Full Changelog: v1.0.1...v1.1.2

TranscribeGeek 1.0.1

Maintenance release.

• Application icon redrawn to match the TechyGeeksHome house style (blue gradient tile, white glyph), so TranscribeGeek now sits correctly alongside the rest of the Geek range.
• No functional changes to transcription.

Downloads below: TranscribeGeekSetup.exe (installer) and TranscribeGeek-portable.exe (no install needed). SHA256SUMS.txt lists the hashes for both.

Full Changelog: v1.0.0...v1.0.1

TranscribeGeek 1.0.0

First release.

TranscribeGeek turns recordings into text on your own machine. Drop in audio or video and it writes out a transcript and, if you want one, a SubRip .srt subtitle file. It runs OpenAI's Whisper speech models locally through whisper.cpp — no account, no server, no upload, no per-minute limit.

What it does

• WAV, MP3, M4A, FLAC, OGG, Opus, WMA, AAC, MP4, MKV, MOV, AVI, WebM
• Plain text transcript beside the source file, with or without timestamps
• Optional .srt subtitle file
• Queue as many files as you like — worked through one at a time
• 22 languages, or automatic detection

What it will not do

• It does not send your recordings anywhere. The only thing it ever downloads is a speech model, from the Models screen, when you ask for one.
• It does not overwrite your files. If a transcript is already there, the next one is numbered.
• It does not bundle ffmpeg. MP3, MP4 and the rest need it to decode; TranscribeGeek runs it as a separate program if it finds one and says so plainly if it does not. Plain 16 kHz mono WAV works either way.

Speech models

Nothing is shipped in the installer. Pick one on the Models screen and it is fetched once and kept.

Model	Download	When to use it
Tiny	78 MB	Checking a file transcribes at all
Base	148 MB	An older machine
Small	488 MB	The usual choice — start here
Medium	1.5 GB	Accents and poor recordings. Slower

Requirements

Windows 10 1809 or later, 64-bit. .NET 8 is included in the build.

---

Free to use, including at work. GPL-3.0. No paid tier, ever.
Part of the TechyGeeksHome range.

Full Changelog: https://github.com/techygeekshome/TranscribeGeek/commits/v1.0.0