Universal Video Extractor

Any page URL to a direct, downloadable video link. Powered by yt-dlp (1000+ supported sites): articles, blog posts, social media, news pages, and video hosts.

📊 What you get

Clean JSON, one record per video that has a direct download URL:

• source_url — the page URL you supplied
• video_url — direct download URL for the selected format
• title — video title
• duration_secs — duration in seconds
• resolution — WIDTHxHEIGHT of the selected format
• format — container of the selected format (mp4, webm, …)
• filesize_bytes — exact or approximate size of the selected format
• thumbnail_url — thumbnail image URL
• extractor — which yt-dlp extractor handled the page
• webpage_url — canonical page URL as resolved by the extractor
• uploader — channel / account name
• upload_date — YYYYMMDD
• metadata — full yt-dlp info dict (omit by turning "Include full metadata" off)

💵 Pricing

• Actor start: $0.00005 per event (one event per GB of run memory, minimum one)
• Result: $0.05 per video link delivered — about 20 videos per $1

One charge per row delivered, and a row is only written when a direct download URL was actually found:

• No urls given → 0 results, nothing charged, and the status message names the field. The Actor has no hidden default URL, so an empty run never does billable work.
• A page with nothing extractable (private, removed, unsupported, region-locked) → no row, no charge; the page is counted in the status message.
• A video with no format matching your formats / maxQuality → no row, no charge.
• Page blocked or the network failed → the Actor retries three times, rotating a fresh residential proxy session each time. If nothing at all could be extracted, the run fails with a status message naming the cause instead of finishing green and empty.
• Run time limit → the Actor stops cleanly, says "stopped early at the run time limit", and you keep (and pay for) only the rows already delivered.

Diagnostics never enter the dataset. Errors, skips and counts go to the run status message and the OUTPUT record in the key-value store.

⚙️ Input

Field	Meaning
urls	Page URLs to extract from (required)
formats	Container preference in priority order; defaults to mp4 then webm
maxQuality	Height cap: 360p…2160p, or best for no cap (default 1080p)
includeMetadata	Attach the full yt-dlp info dict (default on)
maxUrls	Safety cap on how many input URLs one run processes (default 100)

🤖 Use with AI agents

Point Claude, the OpenAI Agents SDK, an n8n flow or any MCP-aware client at it.

Sample agent prompt:

Run Universal Video Extractor on my input and return the structured results.

Agentic payments (x402): Supports agentic payment via x402 — call this actor with USDC, no API key required.

🔗 Related tools — Video tools

YouTube Metadata · Video Link Extractor · YouTube Transcripts

---

A NexGenData utility actor. Download URLs are time-limited by the host — use them promptly. Respect each site's terms and copyright.