SpeakerDeck Scraper

Purpose

Download all slide images from a SpeakerDeck presentation URL. Zero dependencies — uses only built-in fetch and fs APIs. The downloaded images can be read by Claude Vision or converted with markitdown.

Run

nix-shell -p bun --run "bun {baseDir}/scripts/speakerdeck.ts <speakerdeck-url> [-o output-dir]"

Arguments

Arg	Description
<url>	SpeakerDeck presentation URL (required)
-o, --output <dir>	Output directory (default: $TMPDIR/speakerdeck-<id>/)

Output (JSON to stdout)

{
	"title": "Presentation Title",
	"presentationId": "abc123def456",
	"slideCount": 30,
	"outputDir": "/tmp/speakerdeck-abc123def456",
	"files": ["/tmp/speakerdeck-abc123def456/slide_000.jpg", "..."]
}

Workflow Examples

Read slides with Claude Vision

result=$(nix-shell -p bun --run "bun {baseDir}/scripts/speakerdeck.ts 'https://speakerdeck.com/user/talk'")
# Then use the file paths from the JSON output with Read tool

Convert to Markdown with markitdown

result=$(nix-shell -p bun --run "bun {baseDir}/scripts/speakerdeck.ts 'https://speakerdeck.com/user/talk'")
dir=$(echo "$result" | jq -r '.outputDir')
for f in "$dir"/*.jpg; do markitdown "$f"; done

How It Works

1. Fetches the SpeakerDeck page HTML to extract the presentation ID (from the player iframe URL)
2. Binary-search probes the CDN (files.speakerdeck.com) to determine slide count
3. Downloads all slide JPEGs (5 concurrent)

Notes

• Images are 1920x1080 JPEG from SpeakerDeck's CDN
• Zero external dependencies — bun is provided via nix-shell -p bun
• Private or password-protected presentations are not supported