Initial commit of video_extractor toolkit

Widevine/DRM research and video download scripts (Apple, CDA, Sky, TVP), CDM tooling, and helper utilities. Excludes downloads/. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01WFUhtZEpkGZJjNCGFS9r4Z
2026-08-08 12:35:28 +02:00
commit 453ca2f85a
227 changed files with 21028 additions and 0 deletions
@@ -0,0 +1,121 @@
+TVP VOD Downloader
+==================
+
+Downloads Widevine L3 DRM-protected videos from vod.tvp.pl as decrypted MP4 files.
+
+
+REQUIREMENTS
+------------
+Python packages:
+    pip install pywidevine requests
+
+System tools (already present in ./tools/):
+    tools/N_m3u8DL-RE   — DASH segment downloader
+    tools/mp4decrypt     — CENC decryption (Bento4)
+    ffmpeg               — must be in PATH (used for muxing)
+
+Widevine device file (already present in ./L3/):
+    L3/google_sdk_gphone64_x86_64_3198748f_22596_l3.wvd
+
+
+USAGE
+-----
+
+1. Download a single episode by product ID:
+
+    python3 tvp_download.py 2993670
+
+2. Download multiple episodes by product ID:
+
+    python3 tvp_download.py 2726892 2785411 2871666
+
+3. Download all episodes of a series by series ID:
+
+    python3 tvp_download.py --series 2726890
+
+Output files are saved to ./downloads/ as:
+    sXXeXX_title.mp4   (e.g. s01e13_zaslubiny.mp4)
+
+
+FINDING IDs
+-----------
+
+Product ID and series ID come from the vod.tvp.pl URL:
+
+    https://vod.tvp.pl/seriale,18/nazwa-serialu-odcinki,{SERIES_ID}/odcinek-N,S01E01,{PRODUCT_ID}
+
+Example — "Wojna zastępcza":
+    Series ID:  2726890   (from the URL segment "...-odcinki,2726890/...")
+    Episode 13: 2993670   (from the URL segment "...,S01E13,2993670")
+
+You can also pass the URL directly to the browser and copy the product ID from the address bar
+when playing an episode.
+
+
+HOW IT WORKS
+------------
+
+For each episode:
+  1. Fetches the DASH playlist:
+       GET /api/products/{id}/videos/playlist?platform=BROWSER&videoType=MOVIE
+  2. Extracts the Widevine PSSH from the MPD manifest.
+  3. Sends a license challenge to TVP's Widevine license server using the
+     L3 device certificate (.wvd file) and pywidevine.
+  4. Gets the AES-128 content key from the license response.
+  5. Downloads all encrypted video and audio segments via N_m3u8DL-RE.
+  6. Decrypts segments with mp4decrypt using the content key.
+  7. Muxes video + audio into a single MP4 with ffmpeg.
+
+
+OUTPUT
+------
+
+Each episode is saved as a single MP4:
+    ./downloads/{product_id}_{title}.mp4
+
+Quality: best available (typically 1920x1080 H.264, AAC audio, Polish).
+Size: roughly 1.2-1.5 GB per 45-minute episode at full quality.
+
+
+EXAMPLE SESSION
+---------------
+
+    # List all episodes of "Wojna zastępcza" and download them all:
+    python3 tvp_download.py --series 2726890
+
+    # Output:
+    # Found 13 episodes:
+    #   E01 | 2726892 | odc. 1 – Werbunek
+    #   E02 | 2785411 | odc. 2 – Legnica
+    #   ...
+    # ============================================================
+    # Product 2726892: 'odc. 1 – Werbunek' [EPISODE]
+    #   Key: d431b94c...:c3e2e41c...
+    #   Running: N_m3u8DL-RE 2726892_odc__1___Werbunek
+    #   [+] Saved: s01e01_werbunek.mp4 (1248 MB)
+    # ...
+
+
+TROUBLESHOOTING
+---------------
+
+"No .wvd file found":
+    The Widevine device file is missing from ./L3/. The file
+    google_sdk_gphone64_x86_64_3198748f_22596_l3.wvd must be present.
+
+"Playlist fetch failed" / 404:
+    The product ID may be wrong, or the content may have been removed from TVP VOD.
+    Verify the URL on vod.tvp.pl.
+
+"License server 406":
+    Do not add Accept: application/json to the license POST request.
+    The fetch_content_key() function uses only Content-Type: application/octet-stream.
+
+"No CONTENT keys returned":
+    The WVD file may have system_id=0 (broken). The script auto-selects the WVD
+    with a non-zero system_id. If all WVDs have system_id=0, the device file needs
+    to be regenerated.
+
+Download hangs at 0%:
+    Avoid piping N_m3u8DL-RE output through head or other tools — the broken pipe
+    stalls the process. Always let it write to stdout directly or redirect to a file.