videocut CLI

videocut is a command-line tool for talking-head video workflows. It handles transcription, subtitle structuring, human-readable review output, edit application, review UI hosting, and final cutting based on approved delete segments.

Installation

npm install -g @huiqinghuang/videocut-cli

After installation, verify that the command is available:

videocut --help

Requirements

• Node.js 18+
• FFmpeg
• Volcengine speech transcription API key: VOLCENGINE_API_KEY

Example:

export VOLCENGINE_API_KEY="your_api_key"

Commands

videocut transcribe <video> -o <output-dir> [--hotwords <hotwords.txt>]
videocut generate-subtitles <volcengine-result.json> -o <subtitles.json>
videocut generate-readable <subtitles.json> -o <readable.txt>
videocut apply-edits <subtitles.json> <edits.json> -o <subtitles-edited.json>
videocut review-server 8899 --path <project-or-output-root>
videocut cut <video> <delete-segments.json> -o <output-video>

Typical Workflow

# 1. Transcribe the video
videocut transcribe input.mp4 -o output/demo --hotwords hotwords.txt

# 2. Build the subtitle structure
videocut generate-subtitles output/demo/1_transcribe/volcengine_result.json

# 3. Generate readable review text for AI or human review
videocut generate-readable output/demo/common/subtitles_words.json -o output/demo/2_analysis/readable.txt

# 4. Apply edits from edits.json back to the structured subtitles
videocut apply-edits output/demo/common/subtitles_words.json output/demo/2_analysis/edits.json

# 5. Start the review server
videocut review-server 8899 --path output/demo

Hotwords File

Use a plain text file with one term per line:

container_of
offsetof
list_entry
GitHub
Gist

If --hotwords is provided, the list is submitted to Volcengine ASR and also recorded to 1_transcribe/hotwords_used.json.

Related Project

• Skills repository: videocut-skill
• Repository: https://github.com/sunnyswag/videocut