BIG docs refactoring

This commit is contained in:
Alex X
2026-02-06 20:43:50 +03:00
parent 5bf5327a45
commit d01b99d105
67 changed files with 2104 additions and 1916 deletions
+52 -60
View File
@@ -1,70 +1,62 @@
# ffmpeg
# FFmpeg
## FFplay output
You can get any stream, file or device via FFmpeg and push it to go2rtc. The app will automatically start FFmpeg with the proper arguments when someone starts watching the stream.
[FFplay](https://stackoverflow.com/questions/27778678/what-are-mv-fd-aq-vq-sq-and-f-in-a-video-stream) `7.11 A-V: 0.003 fd= 1 aq= 21KB vq= 321KB sq= 0B f=0/0`:
- FFmpeg preinstalled for **Docker** and **Home Assistant add-on** users
- **Home Assistant add-on** users can target files from [/media](https://www.home-assistant.io/more-info/local-media/setup-media/) folder
- `7.11` - master clock, is the time from start of the stream/video
- `A-V` - av_diff, difference between audio and video timestamps
- `fd` - frames dropped
- `aq` - audio queue (0 - no delay)
- `vq` - video queue (0 - no delay)
- `sq` - subtitle queue
- `f` - timestamp error correction rate (Not 100% sure)
## Configuration
`M-V`, `M-A` means video stream only, audio stream only respectively.
## Devices Windows
```
>ffmpeg -hide_banner -f dshow -list_options true -i video="VMware Virtual USB Video Device"
[dshow @ 0000025695e52900] DirectShow video device options (from video devices)
[dshow @ 0000025695e52900] Pin "Record" (alternative pin name "0")
[dshow @ 0000025695e52900] pixel_format=yuyv422 min s=1280x720 fps=1 max s=1280x720 fps=10
[dshow @ 0000025695e52900] pixel_format=yuyv422 min s=1280x720 fps=1 max s=1280x720 fps=10 (tv, bt470bg/bt709/unknown, topleft)
[dshow @ 0000025695e52900] pixel_format=nv12 min s=1280x720 fps=1 max s=1280x720 fps=23
[dshow @ 0000025695e52900] pixel_format=nv12 min s=1280x720 fps=1 max s=1280x720 fps=23 (tv, bt470bg/bt709/unknown, topleft)
```
## Devices Mac
```
% ./ffmpeg -hide_banner -f avfoundation -list_devices true -i ""
[AVFoundation indev @ 0x7f8b1f504d80] AVFoundation video devices:
[AVFoundation indev @ 0x7f8b1f504d80] [0] FaceTime HD Camera
[AVFoundation indev @ 0x7f8b1f504d80] [1] Capture screen 0
[AVFoundation indev @ 0x7f8b1f504d80] AVFoundation audio devices:
[AVFoundation indev @ 0x7f8b1f504d80] [0] Soundflower (2ch)
[AVFoundation indev @ 0x7f8b1f504d80] [1] Built-in Microphone
[AVFoundation indev @ 0x7f8b1f504d80] [2] Soundflower (64ch)
```
## Devices Linux
```
# ffmpeg -hide_banner -f v4l2 -list_formats all -i /dev/video0
[video4linux2,v4l2 @ 0x7f7de7c58bc0] Raw : yuyv422 : YUYV 4:2:2 : 640x480 160x120 176x144 320x176 320x240 352x288 432x240 544x288 640x360 752x416 800x448 800x600 864x480 960x544 960x720 1024x576 1184x656 1280x720 1280x960
[video4linux2,v4l2 @ 0x7f7de7c58bc0] Compressed: mjpeg : Motion-JPEG : 640x480 160x120 176x144 320x176 320x240 352x288 432x240 544x288 640x360 752x416 800x448 800x600 864x480 960x544 960x720 1024x576 1184x656 1280x720 1280x960
```
## TTS
Format: `ffmpeg:{input}#{param1}#{param2}#{param3}`. Examples:
```yaml
streams:
tts: ffmpeg:#input=-readrate 1 -readrate_initial_burst 0.001 -f lavfi -i "flite=text='1 2 3 4 5 6 7 8 9 0'"#audio=pcma
# [FILE] all tracks will be copied without transcoding codecs
file1: ffmpeg:/media/BigBuckBunny.mp4
# [FILE] video will be transcoded to H264, audio will be skipped
file2: ffmpeg:/media/BigBuckBunny.mp4#video=h264
# [FILE] video will be copied, audio will be transcoded to PCMU
file3: ffmpeg:/media/BigBuckBunny.mp4#video=copy#audio=pcmu
# [HLS] video will be copied, audio will be skipped
hls: ffmpeg:https://devstreaming-cdn.apple.com/videos/streaming/examples/bipbop_16x9/gear5/prog_index.m3u8#video=copy
# [MJPEG] video will be transcoded to H264
mjpeg: ffmpeg:http://185.97.122.128/cgi-bin/faststream.jpg#video=h264
# [RTSP] video with rotation, should be transcoded, so select H264
rotate: ffmpeg:rtsp://12345678@192.168.1.123/av_stream/ch0#video=h264#rotate=90
```
## Useful links
All transcoding formats have [built-in templates](ffmpeg.go): `h264`, `h265`, `opus`, `pcmu`, `pcmu/16000`, `pcmu/48000`, `pcma`, `pcma/16000`, `pcma/48000`, `aac`, `aac/16000`.
- https://superuser.com/questions/564402/explanation-of-x264-tune
- https://stackoverflow.com/questions/33624016/why-sliced-thread-affect-so-much-on-realtime-encoding-using-ffmpeg-x264
- https://codec.fandom.com/ru/wiki/X264_-_описание_ключей_кодирования
- https://html5test.com/
- https://trac.ffmpeg.org/wiki/Capture/Webcam
- https://trac.ffmpeg.org/wiki/DirectShow
- https://stackoverflow.com/questions/53207692/libav-mjpeg-encoding-and-huffman-table
- https://github.com/tuupola/esp_video/blob/master/README.md
- https://github.com/leandromoreira/ffmpeg-libav-tutorial
- https://www.reddit.com/user/VeritablePornocopium/comments/okw130/ffmpeg_with_libfdk_aac_for_windows_x64/
- https://slhck.info/video/2017/02/24/vbr-settings.html
- [HomeKit audio samples problem](https://superuser.com/questions/1290996/non-monotonous-dts-with-igndts-flag)
But you can override them via YAML config. You can also add your own formats to the config and use them with source params.
```yaml
ffmpeg:
bin: ffmpeg # path to ffmpeg binary
global: "-hide_banner"
timeout: 5 # default timeout in seconds for rtsp inputs
h264: "-codec:v libx264 -g:v 30 -preset:v superfast -tune:v zerolatency -profile:v main -level:v 4.1"
mycodec: "-any args that supported by ffmpeg..."
myinput: "-fflags nobuffer -flags low_delay -timeout {timeout} -i {input}"
myraw: "-ss 00:00:20"
```
- You can use go2rtc stream name as ffmpeg input (ex. `ffmpeg:camera1#video=h264`)
- You can use `video` and `audio` params multiple times (ex. `#video=copy#audio=copy#audio=pcmu`)
- You can use `rotate` param with `90`, `180`, `270` or `-90` values, important with transcoding (ex. `#video=h264#rotate=90`)
- You can use `width` and/or `height` params, important with transcoding (ex. `#video=h264#width=1280`)
- You can use `drawtext` to add a timestamp (ex. `drawtext=x=2:y=2:fontsize=12:fontcolor=white:box=1:boxcolor=black`)
- This will greatly increase the CPU of the server, even with hardware acceleration
- You can use `timeout` param to set RTSP input timeout in seconds (ex. `#timeout=10`)
- You can use `raw` param for any additional FFmpeg arguments (ex. `#raw=-vf transpose=1`)
- You can use `input` param to override default input template (ex. `#input=rtsp/udp` will change RTSP transport from TCP to UDP+TCP)
- You can use raw input value (ex. `#input=-timeout {timeout} -i {input}`)
- You can add your own input templates
Read more about [hardware acceleration](hardware/README.md).
**PS.** It is recommended to check the available hardware in the WebUI add page.
+22
View File
@@ -0,0 +1,22 @@
# FFmpeg Device
You can get video from any USB camera or Webcam as RTSP or WebRTC stream. This is part of FFmpeg integration.
- check available devices in web interface
- `video_size` and `framerate` must be supported by your camera!
- for Linux supported only video for now
- for macOS you can stream FaceTime camera or whole desktop!
- for macOS important to set right framerate
## Configuration
Format: `ffmpeg:device?{input-params}#{param1}#{param2}#{param3}`
```yaml
streams:
linux_usbcam: ffmpeg:device?video=0&video_size=1280x720#video=h264
windows_webcam: ffmpeg:device?video=0#video=h264
macos_facetime: ffmpeg:device?video=0&audio=1&video_size=1280x720&framerate=30#video=h264#audio=pcma
```
**PS.** It is recommended to check the available devices in the WebUI add page.