feat: Add image and video download support to rn-wechat-extract

[runningZ1]

Description

Adds enhanced versions of fetch_wechat.py that download images, native videos, and WeChat Channels (视频号) videos embedded in WeChat public account articles.

New Scripts

fetch_wechat_with_images.py

• Downloads images from WeChat articles
• Saves images to {title}_images/ directory

fetch_wechat_with_media.py

• Downloads both images and directly-embedded (mpvideo.qpic.cn) videos

fetch_wechat_with_channels.py 🆕

• All features from with_media
• Resolves and downloads WeChat Channels (视频号) videos, which have no directly playable URL in the article HTML at all

Video Channel (视频号) Video Download

A 视频号 video embedded in an article only exposes an export_id (data-id="export/...") and object_nonce_id (data-nonceid) pair inside a <mp-common-videosnap> tag — neither is a usable download URL on its own, and object_nonce_id is not the TikHub object_id (an earlier version of this PR mixed these up).

Resolution pipeline (no manual copy-paste needed anymore):

1. Extract export_id + object_nonce_id from the article HTML (or use a direct weixin.qq.com/sph/... share link if the article already has one).
2. TikHub fetch_video_detail (export_id + object_nonce_id) → resolves the real numeric object_id.
3. TikHub fetch_video_share_url (object_id) → weixin.qq.com/sph/xxx share link.
4. Community worker sph.litao.workers.dev fetch_video_profile (share_url) → an h264/h265 CDN url carrying an X-snsvideoflag query param.
5. Download that CDN url with WeChat headers.

Step 4's X-snsvideoflag param turns out to make WeChat's CDN return the plaintext, pre-transcoded stream (X-encflag: 0 in the response headers) instead of the ISAAC64-encrypted original that TikHub's own media.full_url points to — so this pipeline needs no video decryption step at all, despite 视频号 videos normally requiring WeChat's proprietary WASM-based decryptor to play.

Requires a TikHub API key (--tikhub-key or TIKHUB_API_KEY env var). Steps 2–3 cost $0.01 each ($0.02/video total); step 4 is free but is a third-party community service outside this project's control. Without a key, 视频号 videos are still detected and reported (export_id/object_nonce_id printed to stderr) but not downloaded — everything else still works.

Usage

# Text only (original)
python3 fetch_wechat.py "<url>" --output-dir "./output"

# With images
python3 fetch_wechat_with_images.py "<url>" --output-dir "./output"

# With images and native videos
python3 fetch_wechat_with_media.py "<url>" --output-dir "./output"

# With images, native videos, and 视频号 videos
python3 fetch_wechat_with_channels.py "<url>" --output-dir "./output" --tikhub-key "$TIKHUB_API_KEY"

Testing Results

Verified end-to-end (real network calls, real TikHub key) against 4 live articles, downloaded files checked with ffprobe for valid duration/codec/no corruption:

Article	Images	Native videos	视频号 video	Result
三维图像声呐成果展示	17/17 ✅	8/16 (rest: expired token, unrelated)	1/1 ✅	720p, 1094.7s, h264+aac
公众号文章也能放视频号的视频	7/7 ✅	—	1/1 ✅	576×1024, 17.5s, h264+aac
收藏！世界杯淘汰赛完整对阵图	3/3 ✅	—	1/1 ✅	576×1024, 11.5s, h264+aac
早啊！新闻来了（央视新闻）	11/11 ✅	—	1/1 ✅	720×1280, 131.5s, h264+aac

4/4 视频号 videos resolved and downloaded successfully; all played back correctly and matched their expected duration.

Changes

• Added scripts/fetch_wechat_with_images.py
• Added scripts/fetch_wechat_with_media.py
• Added/rewrote scripts/fetch_wechat_with_channels.py: replaced the old speculative wxv_/bare-finder/bare-nonceid detection with <mp-common-videosnap>-based export_id + object_nonce_id extraction, and a working TikHub + community-worker resolution and download pipeline
• Updated SKILL.md with full documentation for all three enhanced versions, including the 视频号 resolution pipeline and its TikHub cost

Technical Details

Native videos (mpvideo.qpic.cn):

• URLs contain an authentication token that expires; download can fail if the article HTML is stale — re-scrape close to download time (independent of the 视频号 pipeline above)

视频号 videos:

• See resolution pipeline above. data-nonceid in the article HTML is object_nonce_id, not object_id — do not pass it directly to TikHub's fetch_video_share_url
• Both TikHub and the sph.litao.workers.dev worker sit behind Cloudflare and reject Python's default urllib User-Agent (403); the script sends a browser UA

Limitations

• 视频号 downloads require a paid TikHub API key ($0.02/video)
• sph.litao.workers.dev is a third-party community deployment, not an official API — availability isn't guaranteed
• Native video token expiry (pre-existing, unrelated to this update) is still unresolved