WordCut Video Editor

Free to use. No cost, no account, no subscription.

Edit video by editing text. WordCut transcribes speech with Whisper, lets you click words to remove them, and exports a new MP4 with those sections cut out.

Built with Python, pywebview, FFmpeg, and faster-whisper.

Licensed under the MIT License — use, modify, and share freely.

What it does

1. Load a local MP4/MKV/MOV/WebM file or a YouTube URL
2. Transcribe speech with word-level timestamps
3. Click words in the transcript to mark them for removal
4. Export a new video: OriginalName_edited.mp4 (same folder, original untouched)

System requirements

Parameter	Value
OS	Windows 10/11 (64-bit)
RAM	4 GB+ recommended
Disk space	~200 MB for standalone app
WebView2	Microsoft Edge WebView2 Runtime (preinstalled on most PCs)
Internet	Not required after standalone build is downloaded

Quick start (standalone app)

If you have the release zip:

1. Unzip WordCut-Windows.zip
2. Run WordCut.exe
3. Choose a video → Transcribe video → edit words → Export edited MP4

Developer setup

Prerequisites

Parameter	Value
Python	3.11+ (tested on 3.13)
FFmpeg	Optional for dev mode (bundled automatically during build)
Git	For cloning and publishing

Install dependencies

cd PyScript
python -m pip install -r requirements.txt

Run from source

python app.py

Build standalone Windows app

python -m pip install -r requirements-build.txt
python build.py

Outputs:

Output	Path
App folder	dist/WordCut/
Shareable zip	dist/WordCut-Windows.zip

build.py automatically downloads FFmpeg and the Whisper base model into resources/ before packaging.

Usage

Local video

1. Click Choose MP4 file
2. Review the estimated transcription time shown under the file path
3. Click Transcribe video
4. Click words to remove (red = cut from export)
5. Click Export edited MP4

YouTube

1. Paste a URL
2. Click Load YouTube
3. Edit the transcript
4. Export

Cancel

Use Cancel to stop a running task and clear the current file, URL, and transcript.

Export naming

Input	Output
D:\Videos\Virtual Reality.mp4	D:\Videos\Virtual Reality_edited.mp4

The original file is never overwritten.

Configuration parameters

These are the main tunable values in the codebase.

App paths (lib/paths.py)

Parameter	Default	Description
APP_NAME	VideoTextEditor	Folder name under %LOCALAPPDATA%
APP_TITLE	WordCut Video Editor	Window title

User data locations (runtime)

Data	Path
Logs	%LOCALAPPDATA%\VideoTextEditor\debug.log
YouTube downloads	%LOCALAPPDATA%\VideoTextEditor\downloads\
Whisper model (dev mode)	%LOCALAPPDATA%\VideoTextEditor\whisper-models\

Whisper / transcription (lib/transcribe.py, lib/whisper_worker.py)

Parameter	Default	Description
model_size	base	faster-whisper model (tiny, base, small, medium, large-v3)
device	cpu	Inference device
compute_type	int8	Quantization for CPU
word_timestamps	True	Per-word cut timing
vad_filter	False, then True	Retries with voice activity detection if first pass finds nothing
Audio sample rate	16000 Hz	FFmpeg extraction target
Audio channels	1 (mono)	FFmpeg extraction target

Transcription time estimate (lib/video_info.py)

Parameter	Default	Description
overhead	18 seconds	Model load + audio extraction
processing	duration × 0.35	CPU transcription estimate
Minimum estimate	20 seconds	Floor for very short clips

Formula:

estimate_seconds = max(20, round(18 + video_duration_seconds * 0.35))

Video export / cutting (lib/video_edit.py)

Parameter	Default	Description
padding	0.04 seconds	Extra time kept around each word
Video codec	libx264	Re-encode per segment
Video preset	fast	FFmpeg speed/quality tradeoff
Audio codec	aac	Output audio codec
Export suffix	_edited.mp4	Appended before extension

YouTube download (lib/youtube.py)

Parameter	Value
Format	bv*+ba/b (best video + audio)
Merge format	mp4
Output template	%(title).80B-%(id)s.%(ext)s

Build script (build.py)

Parameter	Value
APP_NAME	WordCut
FFMPEG_URL	gyan.dev FFmpeg essentials build
Whisper model bundled	base
Package tool	PyInstaller (folder bundle, not one-file)

Project structure

PyScript/
  app.py                 # Main pywebview app
  build.py               # Standalone Windows build script
  requirements.txt       # Runtime dependencies
  requirements-build.txt # Build dependencies
  lib/
    paths.py             # App paths and frozen/bundled detection
    ffmpeg_util.py       # FFmpeg/FFprobe helpers
    video_info.py        # Duration, audio check, time estimate
    transcribe.py        # Audio extract + Whisper pipeline
    whisper_worker.py    # Whisper subprocess worker
    video_edit.py        # FFmpeg cut + concat export
    youtube.py           # yt-dlp download
    task_manager.py      # Cancel support
  ui/
    index.html
    app.js
    style.css

GitHub upload

This repo is designed to be published without large binaries. resources/, build/, and dist/ are gitignored and recreated by build.py.

First-time publish

cd PyScript
git init
git add .
git commit -m "Initial commit: WordCut video editor"
git branch -M main
git remote add origin https://github.com/YOUR_USERNAME/wordcut.git
git push -u origin main

Recommended GitHub settings

Setting	Recommendation
Release assets	Upload dist/WordCut-Windows.zip as a GitHub Release (not in git)
.gitignore	Already excludes build artifacts and models
License	Add a LICENSE file if you plan to share publicly

Troubleshooting

Problem	Likely cause
No words after transcribe	Video has no clear speech (music-only, silent, very quiet)
App window blank	Install WebView2 Runtime
Transcription slow	Normal on CPU; longer videos take longer
Cancel ignored briefly	Whisper/ffmpeg may take a moment to stop

Check logs at:

%LOCALAPPDATA%\VideoTextEditor\debug.log

License

WordCut is free and open source under the MIT License.

You may:

• Use it for personal or commercial projects
• Share the app with others
• Modify the source code
• Build and redistribute your own version

No payment required. Attribution is appreciated but not required beyond keeping the license notice in source distributions.