Documentation
How to use every feature — and answers to common questions.
How do I get started with Bowdler?
Bowdler is a macOS app that automates repetitive video editing tasks — silence removal, transcript editing, subtitle generation, profanity censorship, and multicam sync — using local AI that runs entirely on your Mac.
- Download and open the app. No installer — just drag Bowdler to Applications and launch it.
- Pick a mode. Choose what you want to do: Transcript Edit, Silence Removal, Subtitles, Censorship, or Multicam.
- Drop your media. Drag a video or audio file into the app. Bowdler accepts MP4, MOV, M4V, MXF, MP3, WAV, AAC, M4A, and AIFF — most codecs that macOS natively supports.
- Choose an AI engine. Select a Whisper or Parakeet model. Larger models are more accurate but slower. You choose which models to download, keep, or remove at any time.
- Click Process. The app transcribes and detects segments automatically. Everything happens on your Mac — no upload needed.
- Review on the timeline. Toggle segments on or off, adjust settings, and preview before exporting.
- Export to your NLE. Choose your preferred format — most modes support FCPXML, XMEML, SRT, VTT, or a flattened media file. DaVinci Resolve timeline export is available in Subtitles mode. Multicam export is FCPXML only.
What Mac do I need to run Bowdler?
macOS 14 or later, Apple Silicon (M1 / M2 / M3 / M4). An Intel build is not available. The download is ~379 MB. Installed size starts at ~1.2 GB (app only), plus speech models you choose to download.
- RAM: 8 GB minimum (16 GB recommended for medium Whisper model).
- Disk space: models range from ~50 MB to 2,951 MB. You choose which models to download, keep, or remove at any time.
- Models are downloaded on first use and cached in
~/Library/Application Support/com.whyang.bowdler/models/.
Is there a free trial?
You get 8 real exports with no watermark and no account required. All editing features unlocked — basic Whisper models only.
UI Overview
Main Screen
The first screen you see when you open Bowdler. From here you select a mode, drop in your media, choose an AI engine, and start processing.
| # | Element | Description |
|---|---|---|
| 1 | Current Mode | The active tab — Transcript Edit, Censorship, Silence Removal, Subtitles, or Multicam. Click to switch. |
| 2 | Settings Button | Opens the settings panel for the current mode. |
| 3 | Theme Button | Toggles between dark and light theme. |
| 4 | Recent Files | Files you've processed before. Click to reopen. |
| 5 | Upload Area | Drag & drop your media here, or click to open a file picker. Accepts MP4, MOV, M4V, MXF, MP3, WAV, AAC, M4A, AIFF. |
| 6 | Current Model | Shows the active AI engine and model size. Click to change. |
| 7 | Process Button | Starts detection and opens the Review screen when done. |
Timeline / Review Screen
After processing, you're taken to the Review screen where every detected segment is shown on a timeline. You decide what to keep, what to remove, and how to export.
| # | Element | Description |
|---|---|---|
| 1 | Back Button | Returns to the main screen. |
| 2 | Video Visibility | Shows or hides the inline video preview. |
| 3 | Timeline | Visual overview of all detected segments. Click anywhere to jump. |
| 4 | Segment Selection | Quickly check All or uncheck None to include/exclude every segment. |
| 5 | Custom Range | Manually add a time range to censor or remove, independent of detection. |
| 6 | Speed Controls | 1x, 1.25x, 1.5x, 2x playback. |
| 7 | Zoom Controls | Zoom in or out on the waveform. |
| 8 | Playback Controls | Play/pause, skip −10s, −1s, +1s, +10s. |
| 9 | Segment Toggle | Checkbox — controls whether this segment is included in export. |
| 10 | Play Segment | Previews just this one segment in isolation. |
| 11 | Detected Word | The word flagged by the model for this segment. |
| 12 | Duration | Start and end timestamp of the detected segment. |
| 13 | Censor Intensity | Per-segment effect strength from 0% to 150%. If adjusted, overrides the global mute % for this segment. |
| 14 | Export Button | Applies processing and saves the result. |
| 15 | Timeline Export | Exports segments as FCPXML or XML (XMEML) for Final Cut or Premiere. FCPXML only in Multicam. |
How does Transcript Edit work?
Transcribes your video and displays the result as an editable document. Every word is clickable — select sentences or individual words to cut, mute, remove filler words and bad takes, or generate subtitles. Everything in one place without switching modes.
Settings
| Setting | Description |
|---|---|
| Speaker labels | Detects multiple speakers and highlights each in a different color. |
| Mixed Audio | Enable if your recording has mixed audio tracks — improves accuracy. |
| Detect filler words | Highlights detected filler words for easy removal. |
| Export folder | Next to the file, Ask every time, or a fixed custom folder. |
| Reset | Resets the mode settings to default. |
Toolbar
Appears when you select text in the transcript:
| Button | Action |
|---|---|
| Cut | Marks selected words for removal on export. |
| B | Bold — marks selected words for subtitle emphasis. |
| I | Italic — marks selected words for subtitle styling. |
| Color dots | Assigns a speaker color for per-speaker subtitles. |
| × | Clears the selection. |
Tools Menu
| Option | Action |
|---|---|
| Remove Bad Takes | Auto-detects false starts and repeated phrases for removal. |
| Restore Silence | Toggles silence gaps back into the transcript. |
| Remove Filler Words | Auto-marks all detected filler words in one pass. |
| Mute Profanity | Auto-detects and mutes profanity across the transcript. |
| Generate Subtitles | Generates subtitles and opens the subtitle editor. |
| Restore All | Undoes all cuts and restores the original transcript. |
| Save / Load Session | Save or resume an editing session from a file. |
How does automatic censorship work?
Detects swear words using AI and mutes or replaces them with a sound. Every detection is shown before export — nothing gets cut without review.
| Setting | Description |
|---|---|
| Censor Type | Silence, Beep, or Custom audio file. |
| Confidence | Higher = fewer false positives. Lower = catches more. |
| Fuzzy | Lower values catch misspellings and transliterations. |
| Global Mute % | How much of each flagged word to mute. 100% = fully muted. |
| Export Dir | Where the processed file is saved. |
| Custom Dictionaries | Add or remove words from built-in profanity lists. |
| Automute / Markers | Export as FCPXML or XML automute timeline or markers. |
How do I remove silence from a video?
Detects quiet pauses using Voice Activity Detection (VAD) and marks them as segments you can review and remove.
| Setting | Description |
|---|---|
| VAD Threshold | Sensitivity of silence detection. Higher = stricter. |
| Min Silence Dur | Minimum duration a pause must last before it's flagged. |
| Speech Pad | Buffer around speech segments so cuts don't feel abrupt. |
| Fix click sound | Short crossfade at each cut point to eliminate audible clicks. |
| Export Dir | Where the processed file is saved. |
| Autocut / Markers | Export as FCPXML or XML autocut timeline or markers. |
How do I generate subtitles?
Transcribes your video and generates subtitle files you can edit directly in the app before exporting.
| Setting | Description |
|---|---|
| Chars per Line | Maximum characters in a single subtitle line. |
| Lines per Sub | 1 or 2 lines per subtitle block. |
| Split at sentences | Starts a new subtitle at . ! ?. Recommended ON. |
| Scene Detection | Forces a subtitle break at each hard cut. |
| One Word per Subtitle | Animated subtitles. |
| Remove Periods | Strips sentence-ending periods from subtitle text. |
| Speaker Dash | Prepends - to every subtitle line. |
| Text Case | Keep original, ALL CAPS, or all lowercase. |
| Max Duration | Maximum display time for a single subtitle block. |
| Min Pause | Minimum gap between consecutive subtitle blocks. |
| Linger | How long the subtitle stays after speech ends. |
| Formats | SRT, VTT, FCPXML, or DaVinci Resolve timeline. |
| FCPXML Settings | Frame rate, minimum gap, and style settings. |
How does multicam sync work?
Import multiple camera angles and external audio. Sync by waveform, transcript, LTC timecode, or clap/slate. Review on a timeline with camera tiles and switch points. Export an edit decision list as FCPXML (Final Cut Pro). Other NLE exports and flattened video export are in development. Save your work as a .bwdr project file.
| Feature | Description |
|---|---|
| Sync | Waveform (MFCC), transcript alignment, LTC timecode, clap/slate, or manual offset. |
| Drift correction | Detects and compensates clock drift on long recordings. |
| Timeline | Camera tiles, transcript panel, switch points, and a bottom track view. |
| Auto-switching | Bind microphones or diarized speakers to camera angles. Switch points are generated automatically. |
| Track control | Rename, reorder, show/hide video or audio per track. Set main audio. |
| Undo / Redo | Full delta-based history. |
| Proxy playback | ProRes Proxy generated for smooth multicam preview. Proxies are cached and reused. |
| Export | FCPXML 1.10 only. Other NLE exports and flattened video export are in development. |
| Project files | Save and load .bwdr files. Relink missing media when moving projects. |
Planned: Native DaVinci Resolve and Adobe Premiere Pro multicam export. Flat video export with baked-in switches. Integration of existing standalone tools into the multicam timeline — Transcript Edit (bad takes, filler words, censorship, subtitles), Silence Removal. Speaker mapping UI. Batch export presets.
What is Whisper and which model should I choose?
A neural speech recognition model that runs entirely on your Mac — no data ever leaves your computer. Used in Transcript Edit, Censorship, and Subtitles modes for high-accuracy transcription across 99+ languages. Optimized for Apple Silicon via MLX.
tiny ~2 GB RAM · Fastest · Low accuracy
base ~3 GB RAM · Fast · Average accuracy
small ~6 GB RAM · Medium · Good accuracy
medium ~10 GB RAM · Slow · Great accuracyTip: Use small or medium for the best balance. Use tiny or base when speed matters more than accuracy.
Benchmark: On a MacBook Air M4 running macOS 15.5, Whisper Small achieves 3.39% WER on a typical interview clip with an RTF of 0.038 (~26× faster than real-time). Source: Bowdler internal benchmark, July 2026.
What is Parakeet and how is it different from Whisper?
A fast, accurate neural speech recognition model from NVIDIA. Runs entirely on your Mac via a local C++ binary (GGUF). Used in Transcript Edit, Censorship, and Subtitles modes. Parakeet models are generally faster than Whisper at similar accuracy levels and require less RAM.
tdt-0.6b-v3-q4_k ~644 MB · Fastest · 25 languages
tdt-0.6b-v3-q8_0 ~897 MB · Fast · 25 languages
tdt-1.1b-q4_k ~1.0 GB · Medium · English-only
tdt-1.1b-q8_0 ~1.5 GB · Slower · English-onlyq4_k = heavier compression, smaller file. q8_0 = lighter compression, larger file.
Note: Parakeet is available for licensed users only. It is not included in the free trial.
Tip: Parakeet is often the best choice for English transcription. Try both engines and pick the one that works best for your content.
Benchmark: Parakeet TDT 0.6B Q4_K on the same M4 Mac delivers 3.17% WER at 0.028× RTF for short content. The larger 1.1B Q4_K model averages 5–6% WER on longer videos. Source: Bowdler internal benchmark, July 2026.
What is Vosk and when should I use it?
Another offline speech recognition engine. Used only in Censorship mode. Vosk models require less RAM than Whisper and can be more accurate for certain languages.
Small Vosk models (~50–150 MB) can be installed directly in the app. Large models (400 MB–2 GB) must be downloaded manually:
- Go to alphacephei.com/vosk/models
- Download the zip for your language (e.g.
vosk-model-ru-0.42) - Unzip — you get a folder named
vosk-model-* - In Bowdler: Censorship → Settings → Models → Vosk → Custom Path → Select that folder
- The model is now active
Note: The folder name must start with vosk-model.
Common Questions
Pricing & Licensing
How much does it cost?
$59 one-time payment. No subscription. No account. All features included. Free updates included with your purchase.
Is there a free trial?
Yes. 8 real exports with no watermark and no account required. Limited to basic Whisper models — buying unlocks the full range.
How does license activation work?
You receive a license key via Gumroad after purchase. Enter it in the app once. The license is tied to one active Mac at a time. To transfer to a new machine, deactivate your old one at the Key Management page first, then enter the key on the new Mac. There is a 30-day cooldown between transfers.
Can I use it commercially?
Yes. A single license covers personal and commercial use. There are no per-project or per-seat restrictions beyond one active Mac at a time.
What is the refund policy?
Email me within 14 days for a full refund. No questions asked.
Privacy & Security
Does it work offline?
Almost everything. Transcription, silence detection, subtitles, censorship, and multicam sync all run locally. The only feature that needs internet is license activation (via Gumroad).
What about my data privacy?
Your video and audio never leave your Mac. We do not collect usage analytics, crash reports, or tracking data from licensed users. See the privacy policy for full details.
What AI engines does it use?
Whisper MLX and Parakeet for transcription, Silero VAD for silence detection, sherpa-onnx for speaker diarization, and sentence-transformers for bad-take detection. All run locally on your Mac.
Languages & Exports
What languages are supported?
Transcription supports any language Whisper can handle. Profanity censorship covers 32 languages.
What export formats are supported?
SRT, VTT, FCPXML 1.10, FCP7 XMEML, DaVinci Resolve timeline, and flattened media file. Multicam export is FCPXML only (Final Cut Pro) — other NLE exports and flattened video export are in development.
Troubleshooting
Model download fails or stalls
Models are downloaded on first use from Hugging Face. If a download fails, check your internet connection, then restart the app. Downloaded models are cached in ~/Library/Application Support/com.whyang.bowdler/models/ and do not need to be re-downloaded.
The app says "Not Enough Disk Space"
Model files range from 71 MB to 2,951 MB. You choose which models to download, keep, or remove at any time. Ensure you have enough free space for the models you select.
Export produces an empty or broken file
Check that your source file is not corrupted and that the export directory has write permissions. For FCPXML export, ensure the frame rate in Bowdler matches your project settings.
Multicam sync places clips incorrectly
Ensure all clips have overlapping audio. Silent clips (gimbal shots) are handled automatically, but if the majority of your footage lacks audio, use the Manual Offset sync method instead of Audio Waveform.
I replaced my Mac and the license does not activate
Deactivate your old machine first at the Key Management page, then enter the key on your new Mac. If you no longer have access to the old machine, email me and I'll reset it.