supadata_skill

Overview

This skill provides direct curl-based access to the Supadata API to extract video transcripts and scrape web content into markdown. It focuses on practical, scriptable commands for YouTube, TikTok, Instagram, X, and Facebook videos, plus site crawling and metadata extraction. Use it when you need reliable transcript text, timestamped segments, or markdown-ready page content for AI pipelines.

How this skill works

Calls to the Supadata REST endpoints use the x-api-key header and a simple curl pattern. You pass a target URL (or a JSON job payload for crawls) and choose modes such as native (existing transcript only) or generate (AI-generated), plus text=true/false for plain text vs timestamped chunks. Results are returned synchronously (200) or asynchronously (202 with jobId) for long-running crawls, and many endpoints return JSON ready for jq or further scripting.

When to use it

• You need a plain-text or timestamped transcript from a YouTube, TikTok, Instagram, X, or Facebook video.
• You want to scrape a web page and receive a markdown-formatted version for downstream AI processing.
• You require video or channel metadata (views, likes, duration, subscriber count) for analytics.
• You need to crawl multiple pages and aggregate content asynchronously for indexing or training data.
• You want to translate existing transcripts into another language for localization.

Best practices

• Set SUPADATA_API_KEY as an environment variable and use bash -c '...' when piping to avoid environment clearing bugs.
• Use mode=native to save credits when existing transcripts are available; use generate only when necessary.
• Always URL-encode URLs with --data-urlencode to avoid malformed requests.
• Handle async responses by polling the job endpoint until status is completed.
• Check availableLangs in the response before requesting translations or a specific language.

Example use cases

• Fetch a YouTube video's plain transcript, save as text, and feed into an LLM summarization pipeline.
• Scrape a product documentation page to markdown for ingestion into a vector database.
• Start a site crawl to collect blog posts, then poll the job until completed and export results.
• Extract timestamped segments from a tutorial video to build chapter markers or highlight reels.
• Translate a popular interview transcript into Chinese for localized publishing.