Video Transcriber
Turn any video into subtitles. Works offline, no internet needed after setup.
Quick Start (for non-technical users)
1. Install the app and ffmpeg (one time)
Open Command Prompt and paste:
winget install ffmpeg
winget install Augchan42.Transcriber
Restart Command Prompt after installing.
> Prefer a plain zip? Grab TranscriberPortable.zip from the
> latest release,
> unzip anywhere, and skip to step 2 — you still need winget install ffmpeg.
2. Run the transcriber
- Installed via winget: open Command Prompt and type
transcriber, then press Enter.
- Portable zip: double-click transcriber.exe inside the unzipped folder.
A window opens:
- Click Browse and pick your video file
- Choose your language (use Cantonese/Chinese (zh) for Cantonese)
- Click Transcribe
- Wait for it to finish — subtitle files appear next to your video
That's it! You get a .srt file (subtitles) and a .txt file (plain text) in the same folder as your video.
> First run takes a bit longer because it downloads the speech model (~150MB). After that it works fully offline.
YouTube Upload (optional)
The app can also upload your compressed video directly to YouTube with
automatic resume if the network drops.
First-time setup — one-time, about 5 minutes
Because this is an open-source app, it ships without Google credentials
baked in. You create your own free Google Cloud project and plug it into
the app. This keeps your upload access under your control and avoids
any shared-credential risk.
Follow the short walkthrough: docs/youtube-setup.md
The first time you click Sign in to YouTube without credentials in
place, the app pops open a dialog with buttons to open the guide and the
secrets/ folder — no hunting around required.
Every upload after that
- Pick the video file (usually the
_compressed.mp4 from the compress step)
- Edit the title (auto-filled from filename)
- Pick privacy: unlisted (default), private, or public
- Click Upload to YouTube
If the upload fails mid-way (bad Wi-Fi, laptop sleep, etc.), just click
Upload to YouTube again — it resumes from where it stopped using a
.resume sidecar file.
> Uploads go to the YouTube channel tied to the Google account you signed
> in with, not some shared account.
For Cantonese Videos
- Set language to Cantonese / Chinese (zh)
- If the video is about a specific topic, pick a Domain for better accuracy:
| Domain | Good for |
|---|
| I-Ching / Metaphysics | 易經, 風水, 八字 lectures |
| Philosophy | 儒釋道, 新儒家 discussions |
| Geopolitics | 時事分析, international affairs |
| Finance | 財經, stocks, markets |
| Crypto | Bitcoin, blockchain |
| Technology / AI | Tech and AI topics |
Troubleshooting
"ffmpeg not found" — Make sure you installed ffmpeg (step 1 above) and restarted Command Prompt.
It's very slow — Change the Model to tiny for faster (but less accurate) results. The default base model is a good balance.
Wrong language — Pick the correct language manually instead of leaving it on Auto-detect.
Subtitles look garbled — The tool retries automatically, but if it still looks wrong, try setting the language explicitly.
Developer Setup
If you want to run from source instead of the .exe:
# Install Python 3.10+ and ffmpeg, then:
pip install -r requirements.txt
# GUI
python gui.py
# Command line
python transcribe.py video.mp4
python transcribe.py video.mp4 -l zh --domain iching
python transcribe.py "https://youtube.com/watch?v=..."
Command Line Options
python transcribe.py INPUT [options]
-o DIR Output directory
-m MODEL tiny|base|small|medium|large-v3 (default: base)
-l LANG en, zh (Cantonese), ja, ko, fr, es, de...
--domain DOMAIN iching, geopolitics, finance, philosophy, etc.
--format FORMAT srt|txt|both (default: both)
-v Verbose output
Building the .exe
pip install pyinstaller
python build_exe.py
Produces dist/transcriber.exe (GUI) and dist/transcribe.exe (CLI).