Solid-State Society

Tag: FFmpeg

  • Download Bilibili 4K Videos. The Hard Way

    Requirements:

    • Google Chrome installed
    • Get cookies.txt extension installed
    • youtube-dl installed
    • FFmpeg installed
    • Active bilibili membership (to get 4K streaming perk)
    • Basic JSON and Bash knowledge

    Steps:

    • Open the video page you want to download with Google Chrome
    • Open Web Inspector and refresh the page
    • Change video quality to some lower ones (like 480p) and change back to 4K
    • Search for player/playurl in Inspector
    • You should get a JSON payload returned:

    Then select one of the most recent requests with accept_quality that has 120 in array:

    Then select down to the data.dash.video[0].baseUrl and data.dash.audio[0].baseUrl. Remember these URLs:

    Open Get cookies.txt extension and download your current cookies.

    Download video and audio with youtube-dl:

    youtube-dl --cookies bilibili.com_cookies.txt --referer 'https://www.bilibili.com/video/BV1d5411g7g2' 'https://xxx.mcdn.bilivideo.cn:4483/upgcxcode/12/67/336676712/336676712-1-30232.m4s...'

    Then you will get two files like the following:

    • 20210510 336676712-1-30280 [336676712-1-30280].m4s
    • 20210510 336676712-1-30120 [336676712-1-30120].m4s

    Rename the video extension to mp4 and audio to m4a. Play them to check if everything works.

    Combine these two with FFmpeg:

    ffmpeg -i 20210510\ 336676712-1-30120\ \[336676712-1-30120\].mp4 -i 20210510\ 336676712-1-30280\ \[336676712-1-30280\].m4a -c copy output-combined.mp4

  • Low Latency HomeKit Camera Support: Homebridge FFmpeg Plugin with Custom Codec Configuration

    Stream passthrough method for low latency HomeKit camera support:

    • Device: EZVIZ (萤石) CS-C6CN-3H2CWF (星光夜视版)
    • Low latency (as low as 100ms) from the camera to homebridge-camera-ffmpeg
    • Low CPU load since all streams are passthrough from FFmpeg without re-encoding
    • No audio support due to compatibility issue from the source stream

    The configuration should be look like:

    "videoConfig": {
    "source": "-rtsp_transport tcp -i rtsp://admin:<device_passcode>@<device_ip>:554/h265/ch1/main/av_stream",
    "maxFPS": 30,
    "audio": false,
    "vcodec": "copy",
}

    Here’s the tricky part: the EZVIZ CS-C6CN-3H2CWF will output h265 by default, which is not supported by HAP. And the default bitrate is also too high to handle for iOS devices. So you will need to update some settings for the camera:

    • Download the Chinese version of EZVIZ Studio from the official website (direct link).
    • Install and run
    • Click 本地设备 (Local Devices) on the sidebar
    • Right-click the camera you want to modify, then choose 高级设置 (Advanced Settings)
    • In the popup window, click 图像 (Images) – 视音频 (Audio/Video) on the sidebar, change the following settings:
      • 编码类型 (Encoding Type): STD_H264
      • 码率上限 (Bitrate Limit): 1024 Kbps
      • (Optional) 分辨率 (Resolution): HD720P
    • Click 保存 (Save) to save the settings

    Please note this step will slightly reduce the quality of your video stream when you view the stream in EZVIZ official app, but it’s mandatory for the camera to work with HomeKit.

    Bonus: audio support without compiling FFmpeg after version 1.0.0:

    "videoConfig": {
    "source": "-rtsp_transport tcp -i rtsp://admin:<device_passcode>@<device_ip>:554/h265/ch1/main/av_stream",
    "maxFPS": 30,
    "audio": true
}

    This will make FFmpeg re-encode your stream and hurt your CPU. The latency is about 0.6-1 second. I suggest you install Homebridge on a Mac mini for better performance.

    Update for homebridge-camera-ffmpeg 3.0.0 and later: It’s finally stable enough to stream passthough with audio support. You can now use the following config:

    "videoConfig": {
    "source": "-rtsp_transport tcp -i rtsp://admin:<device_passcode>@<device_ip>:554/h265/ch1/main/av_stream",
    "maxFPS": 30,
    "audio": true,
    "vcodec": "copy",
}

  • How to Use FFmpeg to Extract Audio from Video File Without Re-encoding

    Try -c copy:

    $ ffmpeg -i video.mp4 -c copy audio.aac

    How ever you have to know the original audio format to extract audio without re-encoding, for example:

    $ ffmpeg -i video.mp4
...
  Metadata:
    major_brand     : isom
    minor_version   : 1
    compatible_brands: isomavc1
    creation_time   : 2015-01-17 20:39:33
  Duration: 00:23:59.94, start: 0.000000, bitrate: 1983 kb/s
    Stream #0:0(und): Video: h264 (High) (avc1 / 0x31637661), yuv420p(tv, bt709), 1280x720 [SAR 1:1 DAR 16:9], 1787 kb/s, 23.98 fps, 23.98 tbr, 24k tbn, 47.95 tbc (default)
    Metadata:
      creation_time   : 2015-01-17 20:39:33
      handler_name    : 264:[email protected]
    Stream #0:1(jpn): Audio: aac (LC) (mp4a / 0x6134706D), 48000 Hz, stereo, fltp, 191 kb/s (default)
    Metadata:
      creation_time   : 2015-01-17 15:55:11
      handler_name    : Sound Media Handler
...

    Then you should know what format and which track to extract.