A sophisticated automated pipeline for generating, correcting, and translating YouTube interview transcripts with precision speaker attribution.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install youtube-transcript-pipeline
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 youtube-transcript-pipeline 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 YouTube Transcript Pipeline?
The YouTube Transcript Pipeline is a high-fidelity skill designed for developers and researchers who need structured, reusable data from video content. By leveraging Deepgram Nova-3 and conservative diarization correction passes, it ensures that transcripts are not just raw text, but clean, professional assets ready for handoff or integration into larger AI workflows.
This specific entry in the Openclaw Skills ecosystem specializes in handling the nuances of interview dynamics, such as distinguishing between main speakers and interviewers. It provides a robust framework for managing audio chunking, word-level reconciliation, and optional bilingual output while maintaining strict timestamp integrity across all file versions.
YouTube Transcript Pipeline Use Cases
- Creating professional interview transcript packages for content distribution or archival.
- Correcting inaccurate speaker diarization from raw AI transcription services.
- Generating timestamp-preserved Chinese translations for international audiences.
- Structuring messy video data into organized, portable workbench folders with clear manifests.
How YouTube Transcript Pipeline Works
- Asset Preparation: Downloads audio from a provided YouTube URL and segments it into 180-second chunks for optimized processing.
- Transcription: Executes chunked transcription via Deepgram Nova-3 to obtain high-accuracy, word-level outputs.
- Base Production: Generates baseline JSON and text artifacts, ensuring all intermediate files are preserved for auditing.
- Speaker Correction: Applies a conservative attribution pass to fix diarization, prioritizing the main speaker while identifying interviewer interjections.
- Format-Preserved Translation: Optionally translates the transcript (defaulting to Chinese) while keeping timestamps and speaker tags perfectly aligned.
- Final Packaging: Aggregates all transcripts, audio artifacts, and scripts into a structured workbench folder with a MANIFEST.txt file.
YouTube Transcript Pipeline Setup
To deploy this skill within your workspace, ensure you have your Deepgram API key ready and follow these steps:
# Install the necessary transcription and translation utilities
pip install deepgram-sdk google-cloud-translate
# Run the transcription pipeline for a specific audio directory
python3 resilient_dg_transcribe_resume.py \
--audio-dir /tmp/dg_chunks_work2 \
--max 0 --chunk-sec 180 --model nova-3 --timeout 240
# Initialize the standardized workbench structure
bash create_youtube_transcript_workbench.sh project_name
YouTube Transcript Pipeline Data Schema & Taxonomy
The skill organizes output into a strict hierarchy to ensure compatibility with other Openclaw Skills:
| Directory | Content Description |
|---|---|
transcripts/ |
Contains raw, cleaned, and translated versions of the text. |
artifacts/ |
Stores the original audio chunks and video thumbnails. |
scripts/ |
Workflow helpers like translate_transcript_google.py and rebuild_from_words.py. |
snippets/ |
Small JSON helper files and metadata fragments. |
MANIFEST.txt |
A root-level index of all files included in the package. |
YouTube Transcript Pipeline Advanced Features
- Conservative Attribution Logic: Prevents over-attribution of lines to the interviewer by using a 'main speaker default' bias during the correction pass.
- Bilingual Synchronization: Advanced translation scripts that preserve the
[$TS] Speaker: textformat exactly for easy side-by-side comparison. - Resilient Resumption: Transcription scripts support chunk-based resuming, allowing the pipeline to recover from network timeouts without re-processing the entire video.
- Git-Ready Reproducibility: Includes options for initializing a git repository within the output folder to track changes and audit transcript edits.
SKILL.md
Loading
Related Openclaw Skills
OpenCode ACP Collaboration
A high-performance collaboration skill that leverages the Agent Client Protocol (ACP) to automate complex code modifications and refactoring via OpenCode.
Xiaohu WeChat Format
A powerful utility for converting Markdown into perfectly formatted, inline-styled WeChat Official Account articles with automated media handling and draft publishing.
memories-cli
A powerful command-line interface for managing persistent memory, generating AI tool configurations, and syncing context across development environments.
kmi-cli
A command-line tool for fetching Belgian weather data, forecasts, and warnings directly from the Royal Meteorological Institute (KMI/IRM).
Moltbook
Moltbook is a specialized social networking protocol designed specifically for AI agents to post content, engage in discussions, and form communities.
ZTPC Spam Sweep
An intelligent browser automation skill that scans Aliyun Enterprise Mail for unread spam and phishing attempts using custom heuristics.
Featured*
