Transcriber is a desktop graphical user interface (GUI) designed to convert video and audio files into .srt subtitle files and plain text (.txt) using the faster-whisper speech-to-text engine. It operates fully offline after the initial model download, making it ideal for users who need reliable transcription without internet access.
Key Features:
Multilingual Support: Transcribes in multiple languages, including Cantonese/Chinese (zh), with auto-detection available.
Domain Hints: Enhances accuracy for specialized content such as finance, philosophy, geopolitics, I-Ching, and technology/AI.
Video Compression: Compresses videos to H.264 format using bundled ffmpeg.exe and ffprobe.exe, eliminating the need for additional installations.
YouTube Upload: Supports direct video uploads to YouTube with resumable upload functionality, requiring only user-provided Google OAuth credentials.
Offline Functionality: After the first model download, it works entirely offline, ensuring privacy and reliability.
Audience & Benefit:
Ideal for content creators, educators, YouTubers, podcasters, and researchers who need accurate subtitles or plain text transcripts quickly. Its support for Cantonese/Chinese and domain-specific hints makes it particularly useful for niche content creation. The bundled ffmpeg tools simplify video compression for YouTube uploads, while the offline functionality ensures uninterrupted use even without internet access.
Transcriber can be installed via winget, making setup straightforward for users familiar with package managers.
README
Video Transcriber
Turn any video into subtitles. Works offline, no internet needed after setup.
> 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.
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)
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: