An offline, cross-platform video transcription and subtitle extractor powered by whisper-cpp and OpenAI Whisper.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install guipi-video-transcript
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install guipi-video-transcript using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Video Transcript & Subtitle Extractor?
The Video Transcript & Subtitle Extractor is a robust, local-first utility designed to help users transcribe video files and generate perfectly timed subtitles without relying on cloud services. As one of the most practical Openclaw Skills for content creators, it provides fully offline speech-to-text processing on macOS, Windows, and Linux. By using this tool over external APIs, users get fully confidential, zero-cost processing. It stands out in the directory of Openclaw Skills by combining extreme performance with developer-friendly customizations.
With deep integration of local transcription engines, this tool delivers optimized GPU acceleration on Apple Silicon via whisper-cpp and robust compatibility across other systems via openai-whisper. This ensures complete data privacy and rapid speech-to-text transcription, making it an essential asset for any local AI agent workflow.
Video Transcript & Subtitle Extractor Use Cases
- Extracting speech transcripts from local video files (MP4, MOV, AVI, etc.) to compile into written blogs or social media copy.
- Generating timestamped SRT subtitle files to import directly into video editors like CapCut, Premiere Pro, or DaVinci Resolve.
- Transcribing bilingual or mixed-language video content using Whisper's robust auto-detection algorithm.
- Processing secure, confidential, or private video footage entirely locally, ensuring zero data is transmitted over the internet.
How Video Transcript & Subtitle Extractor Works
- Prerequisite & Path Verification: The system verifies the existence of the source video path and checks if
ffmpegand Python are installed on the local system. - Engine Auto-Detection: Based on the operating system, the script identifies the best transcription engine. It prioritizes whisper-cpp with Apple Silicon Metal GPU acceleration on macOS, and falls back to openai-whisper via pip on Windows or Linux if whisper-cpp is unavailable.
- High-Fidelity Audio Extraction: Using ffmpeg, the skill runs
extract_audio.pyto strip the audio stream and convert it into a 16kHz mono WAV file, the optimal sample rate required by Whisper engines. - Local Subtitle & Transcript Generation: Runs
transcribe.pywith custom user parameters, supporting different model sizes (tiny, base, small) and language constraints (auto-detect, English, Chinese, Japanese, etc.). - Post-Transcription Polishing: Implements automated parsing rules to clean up raw Whisper outputs—converting Traditional Chinese to Simplified Chinese, correcting common homonyms, and adding proper punctuation and sentence splits.
- Result Saving: Saves the compiled pure text (
.txt), subtitled tracks (.srt), and audio output (.wav) in the destination folder.
Video Transcript & Subtitle Extractor Setup
Verify that you have ffmpeg and python3 installed on your platform. You can install system-wide dependencies depending on your operating system:
Platform Installation
macOS (Recommended for Apple Silicon Metal GPU Acceleration)
brew install ffmpeg whisper-cpp
Windows
winget install ffmpeg
Linux
sudo apt install ffmpeg
Run the Pipeline
Trigger the video extraction pipeline with a single command:
python3 ~/.workbuddy/skills/video-transcript/scripts/pipeline.py "/path/to/video.mp4" "/path/to/output_dir" base auto
Parameter Options:
video_path(Required): Absolute path to the source video file.output_dir(Required): Path where output files will be saved.model_size(Optional):tiny(77MB),base(147MB, default recommended), orsmall(465MB for high accuracy).language(Optional):auto(auto-detects, default),zh(Chinese),en(English),ja(Japanese).
Video Transcript & Subtitle Extractor Data Schema & Taxonomy
This skill generates structured local artifacts for every video processed. The generated directory layout is organized as follows:
| File Name | Format | Description / Use Case |
|---|---|---|
{video_name}.txt |
Plain Text | Pure transcription without timestamps; ideal for text editing, summaries, and copywriting. |
{video_name}.srt |
SRT Subtitles | Standard subtitle format containing exact millisecond timestamps; importable into editing software (Premiere Pro, CapCut, DaVinci). |
{video_name}.wav |
16kHz Mono Audio | Extracted audio asset, preserved for secondary audio processing. |
.state.json |
JSON Config | Local configuration state, keeping track of user preferences and author promo opt-out states. |
Video Transcript & Subtitle Extractor Advanced Features
- Automated Engine Switcher: Detects hardware and OS to toggle between native C++ binaries (
whisper-cppwith Apple Silicon Metal acceleration) and Python implementations (openai-whisper). - Granular Execution Scripts: Allows developers to separate audio extraction (
extract_audio.py) and translation transcript processing (transcribe.py) into independent steps. - Dynamic Model Hot-swapping: Switch easily between light footprint models (
tinyfor rapid turnaround) and heavy-duty models (smallfor maximum spelling accuracy). - State-Controlled Opt-out: Integrates developer workflow states through a local
.state.jsonfile, allowing seamless toggling of promotional footers during generation tasks.
SKILL.md
Loading
Related Openclaw Skills
Skill Factory v2.0 (技能创建器)
An automated developer assistant to design, safety audit, package, and publish high-performance Openclaw Skills.
PR Autocheck
An automated post-submit pipeline that analyzes git pull request diffs, executes service health checks, and posts interactive, color-coded report cards to Discord.
Repomix-rs Explorer
An AI-optimized codebase packer and analyzer that condenses local or remote repositories into single, token-efficient files for seamless context loading.
Product Selection Assistant
The Product Selection Assistant automates multi-dimensional e-commerce market research, scoring potential products across 6 critical dimensions and rendering interactive HTML reports.
LLM Token Cost Compressor
An intelligent utility that compresses prompts and context data before sending them to LLMs, slashing token consumption by up to 95% while keeping response quality intact.
Insurance Liability Analyzer
An AI-powered agent skill that extracts and analyzes key data from insurance policies, text, or URLs across eight dimensions to generate interactive HTML visualization reports.
Featured*
