Download Koushare Lectures

Download Koushare Lectures (and other EdgeOne-walled Tencent-VOD video)

Core insight

The web page is behind a Tencent EdgeOne bot wall (a JS cookie challenge, then a CAPTCHA). The video stream is not — it lives on a Tencent VOD CDN (video-play.koushare.com / *.myqcloud.com) protected only by a signed token in the URL. So don't fight the wall: get the signed CDN .m3u8 URL by legitimate means, then download it with yt-dlp/ffmpeg. The stream is standard AES-128 HLS ("SimpleAES") with a fetchable key URI, which yt-dlp and ffmpeg decrypt natively.

This is not a DRM bypass. Tencent VOD's voddrm.token "SimpleAES" is HLS clientside encryption (a normal #EXT-X-KEY:METHOD=AES-128 with a key served over HTTPS), not a license-server DRM like Widevine/FairPlay. Any HLS client that plays it can also save it. (Real DRM — Widevine/PlayReady/FairPlay — is a different beast and is out of scope; if you see a license-acquisition request to a DRM server and encrypted keys you can't fetch, stop.)

When this applies (symptoms)

• curl on any koushare.com path → HTTP 200 but only ~1KB of obfuscated JS that sets EO_Bot_Ssid / __tst_status cookies and reloads.
• Page shows "Security Verification — please check the box below" (TencentEOCaptchaWidget from captcha.eo.gtimg.com).
• yt-dlp <page-url> → ERROR: Unsupported URL (no koushare extractor exists — 1700+ extractors, none match).
• The real stream URL looks like https://video-play.koushare.com/.../voddrm.token.<JWT>.video_*.m3u8?sign=...&t=...&us=koushare2020.

Do NOT waste time on these — they dead-end

• ❌ Scraping the page or guessing API endpoints with curl. Every koushare.com path is EdgeOne-walled.
• ❌ Computing/forging the EO_Bot_Ssid cookie challenge. Even if you pass it, layer 2 is a Tencent CAPTCHA you cannot (and must not) solve. It's a dead end, not a shortcut.
• ❌ Hunting for a yt-dlp koushare extractor. There isn't one.

Procedure

0. Ensure tools (once)

yt-dlp --version || pip3 install -q yt-dlp
ffmpeg -version | head -1 || brew install ffmpeg   # yt-dlp needs ffmpeg to mux HLS → mp4

1. Get the signed CDN stream URL

The page's video backend is a POST to https://api-core.koushare.com/live/v2/live/playbackV2; its JSON response holds the CDN URL. You need a browser that has passed the CAPTCHA. Two ways — pick by whether the Chrome extension is connected:

Method A — user hands you the URL (no browser tooling needed): Ask the user to:

1. Open the lecture page in Chrome and click the verification checkbox so the video plays.
2. Press F12 → Network tab → type m3u8 (or mp4) in the filter box.
3. Play the video for a second — a .m3u8 request appears.
4. Right-click it → Copy → Copy as cURL, and paste that to you (it carries the URL and any cookies). Pasting just the .m3u8 URL also works.

Method B — drive the user's Chrome (if the claude-in-chrome extension is connected):

1. tabs_context_mcp, create a tab, navigate to the page.
2. Ask the user to click the one "Security Verification" checkbox — you must not click it yourself (CAPTCHA/bot-detection is off-limits).
3. Reload the page, then read_network_requests filtered by m3u8. The media request fires early, so a reload after network tracking is active is what surfaces it. Fallbacks: read the playbackV2 POST response body, or read_page for the <video> element's src. (The tab title after the CAPTCHA is the real lecture name — grab it for the filename.)

Either way you end up with a URL like: https://video-play.koushare.com/.../voddrm.token.<JWT>.video_1520228_0.m3u8?sign=...&t=...&us=koushare2020

Check the expiry. The middle segment of voddrm.token.<JWT> is base64 JSON with expireTimeStamp — typically ~12h from issue. Download promptly; the URL dies after that, but the downloaded file is permanent. To read it:

python3 -c 'import base64,json,sys; t=sys.argv[1]; t+="="*(-len(t)%4); print(json.loads(base64.urlsafe_b64decode(t)))' '<middle-token-segment>'

2. Download + decrypt

The URLs are long and full of ~ and & — write the URL to a file and use yt-dlp -a to avoid shell mangling. yt-dlp fetches fragments in parallel and decrypts the AES-128 stream natively:

printf '%s\n' 'PASTE_M3U8_URL_HERE' > url.txt
yt-dlp --no-playlist --concurrent-fragments 8 \
  --retries 10 --fragment-retries 20 \
  --user-agent "Mozilla/5.0" \
  -o "$HOME/Downloads/koushare_lecture.mp4" \
  -a url.txt

ffmpeg alternative (also follows the #EXT-X-KEY URI and decrypts): ffmpeg -user_agent "Mozilla/5.0" -i "URL" -c copy out.mp4

3. Verify

ffprobe -v error -show_entries format=duration -of csv=p=0 out.mp4 \
  | awk '{printf "%d min %d sec\n",$1/60,$1%60}'
ffprobe -v error -show_entries stream=codec_type,codec_name -of csv out.mp4

Confirm the duration matches the lecture and there's both a video (h264) and audio (aac) stream. Name the file from the real lecture title, not the numeric id.

Quick reference

Common mistakes