Electron and React
The macOS desktop shell manages files and local processes, while the React interface handles queues, progress, review, and correction.
Open source and local-first
Find the words buried in your recordings.
VerbaDock turns batches of audio and video into searchable transcripts on your Mac, then gives you practical tools to correct speakers, timing, and text.
Intel Macs (x86_64). Unsigned community build. GNU GPLv3.
Local processingRecordings stay on your Mac during transcription.
Batch workflowQueue files and process them one at a time.
Built for correctionFix speakers, boundaries, and transcript text.
A complete local workflow
From a folder of recordings to useful text.
VerbaDock handles the repetitive work, then keeps you in control when automated speaker detection or transcription needs a human correction.
- 01
Add your recordings
Drop audio or video files into a queue and choose where the results should be saved.
- 02
Transcribe locally
Files are processed one at a time and written as TXT, SRT, WebVTT, TSV, and JSON.
- 03
Review and export
Correct the saved transcript, then export styled HTML, speaker-labelled SRT, or subtitle SRT without speaker names.
The transcript is not a dead end
Correct the details that matter.
Automated transcripts are a starting point. VerbaDock keeps the result editable so names and dialogue can match what was actually said.
- Rename speakers and reassign individual segments
- Split or merge segment boundaries
- Edit transcript text while listening to the source
- Reopen completed work and continue later
- Export corrected results as HTML or SRT
Detected
SPEAKER_01 I think that was in the second session.
Corrected
Mary I think that was in the second session.
Under the hood
A local pipeline from media to editable text.
VerbaDock combines established open-source tools in a desktop workflow. Media preparation, speech recognition, speaker grouping, and result writing run on your Mac.
Bundled FFmpeg
Audio and video inputs are converted into a consistent local audio stream before transcription and playback.
Whisper via faster-whisper
Whisper runs locally through faster-whisper with voice activity filtering and word-level timestamps for accurate review and export.
SpeechBrain and scikit-learn
ECAPA speaker embeddings and local clustering assign generic speaker labels that you can rename, split, merge, or correct.
The current packaged release is built for Intel Macs only. Apple Silicon or other custom builds can be requested through GitHub Issues.
Made for recordings worth revisiting
Search, verify, quote, and study.
Students
Turn lectures and course recordings into text you can search while revising.
Researchers
Review interviews locally and correct speaker identities before analysis.
Journalists
Navigate recorded conversations and verify wording against the original audio.
Creators
Build editable transcripts and subtitle files from long-form recordings.
Professionals
Keep sensitive working material on your own machine during transcription.
VerbaDock 1.0.0
Download the Intel Mac build.
The release page includes the packaged DMG, installation guidance, and checksums. This build is unsigned and not notarized because the project is not enrolled in the Apple Developer Program.
First launch: macOS may block the unsigned app. Try to open VerbaDock once, then use System Settings > Privacy & Security > Open Anyway if you trust the download. Read Apple's guidance.