Skip to content

MJPEG Server implements MJPEG over HTTP using FFmpeg or any other input source capable of piping a multipart JPEG stream to stdout. Its primary use case is providing Webdriver screen recordings.

master
Switch branches/tags
Code

Latest commit

 

Git stats

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
 
 
bin
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

MJPEG Server

MJPEG Server implements MJPEG over HTTP using FFmpeg or any other input source capable of piping a multipart JPEG stream to stdout.

Its primary use case is providing screen recordings for remote Webdriver tests, e.g. for blueimp/wdio.

It is resource efficient and by default only starts the screen recording process if there is at least one HTTP client connected to the server, stopping the recording when there are no more open client connections.

Installation

The MJPEG Server binary can be downloaded for Linux, MacOS and Windows from the releases page or built from source via go get:

go get github.com/blueimp/mjpeg-server

The screencast examples also require FFmpeg to be installed.

Usage

By default, mjpeg-server listens on port 9000 on all interfaces and starts the given recording command when the first HTTP client connects:

mjpeg-server [options] [--] command [args]

The command must write a multipart JPEG stream to stdout, e.g.:

--ffmpeg
Content-Type: image/jpeg

[IMAGE_DATA]
--ffmpeg
Content-Type: image/jpeg

[IMAGE_DATA]
...
--ffmpeg--

A sample multipart JPEG generation program can be tested by running the following command and opening http://localhost:9000 in your browser:

go run main.go -a 127.0.0.1:9000 -- go run mpjpeg/main.go -- gopher.jpg

It simply streams the provided JPEG images in an endless loop.

Options

Available MJPEG Server options can be listed the following way:

mjpeg-server -h
Usage of mjpeg-server:
  -a string
    	TCP listen address (default ":9000")
  -b string
    	Multipart boundary (default "ffmpeg")
  -d	Start command directly
  -p string
    	URL path (default "/")
  -v	Output version and exit

The -d option starts the given recording command directly on initialization of the MJPEG server and keeps it running independently of the number of connected HTTP clients, until the MJPEG server process is stopped.

Screencast

Linux

Start mjpeg-server using the x11grab device, selecting the X11 display via -i :DISPLAY and the matching screen resolution via -video_size WIDTHxHEIGHT option:

mjpeg-server -- ffmpeg \
  -loglevel fatal \
  -probesize 32 \
  -fpsprobesize 0 \
  -analyzeduration 0 \
  -fflags nobuffer \
  -f x11grab \
  -r 15 \
  -video_size 1440x900 \
  -i :0 \
  -f mpjpeg \
  -q 2 \
  -

MacOS

List the available avfoundation input devices:

ffmpeg -f avfoundation -list_devices true -i -

Start mjpeg-server, selecting the capture device via -i INDEX option:

mjpeg-server -- ffmpeg \
  -loglevel error \
  -probesize 32 \
  -fpsprobesize 0 \
  -analyzeduration 0 \
  -fflags nobuffer \
  -f avfoundation \
  -capture_cursor 1 \
  -r 15 \
  -pixel_format yuyv422 \
  -i 1 \
  -f mpjpeg \
  -q 2 \
  -

Windows

On Windows, we can use the built-in gdigrab input device to capture the whole desktop.

Start MJPEGServer using the following command in a Powershell console:

MJPEGServer -- ffmpeg `
  -loglevel error `
  -probesize 32 `
  -fpsprobesize 0 `
  -analyzeduration 0 `
  -fflags nobuffer `
  -f gdigrab `
  -r 15 `
  -i desktop `
  -f mpjpeg `
  -q 2 `
  -

License

Released under the MIT license.

Attributions

The Gopher image used for the tests was designed by Renee French.
The design is licensed under the Creative Commons 3.0 Attributions license.

Gopher

Author

Sebastian Tschan

About

MJPEG Server implements MJPEG over HTTP using FFmpeg or any other input source capable of piping a multipart JPEG stream to stdout. Its primary use case is providing Webdriver screen recordings.

Resources

License

Stars

Watchers

Forks

Sponsor this project

 

Packages

No packages published