Flussonic Media Server Documentation

Contents

Publishing Video to the Server

Flussonic Media Server can accept video from external systems and devices that initiate broadcast. This is called publishing to Flussonic.

Publishing can be used in a situation where the external system has no static IP or where it is located behind the firewall in a private IP network. In this case, Flussonic has no way of directly sending a request for video.

What we call publishing to Flussonic:

  • Transmitting video from a mobile device to Flussonic.
  • Transmitting video from OBS (Open Broadcaster Software) or vMix to Flussonic. Learn more
  • Transmitting video from a webpage to Flussonic via WebRTC. Learn more

What we don't call publishing to Flussonic:

  • Receiving a multicast
  • Ingesting a stream from some source

In those cases, Flussonic Media Server has to connect to the data source. Whereas the case where the connection is not initiated by Flussonic itself is called publishing. For example, publishing is when a mobile device connects to Flussonic to transmit video.

Publishing video to social networks is not publishing to Flussonic and therefore it does not meet Flussonic's definition of publishing as used in this documentation.

Supported protocols

Flussonic Media Server can receive requests for video publishing via the RTMP, RTSP, HTTP MPEG-TS, WebRTC, and SRT (20.08) protocols.

Contents:

Publishing to a static stream

If you know what stream name an external system will use to publish video to Flussonic, you can create a stream with that name and a url publish:// data source.

stream published {
  url publish://;
}

To apply the settings, run the command:

/etc/init.d/flussonic reload

Important! In Flussonic version 19.01 and higher, the url publish:// data source type replaces the publish_enabled option. Newer versions are not compatible with older configuration files.

The new url publish:// data source offers far more flexibility - it adds options for timeout, transcoding, and WebRTC support.

Learn more about how to set up publishing to a static stream

URLs for publishing via different protocols

You can publish videos to Flussonic using the following URLs:

  • rtsp://flussonic-ip/published
  • http://flussonic-ip/published/mpegts
  • rtmp://flussonic-ip/published
  • rtmp://flussonic-ip/static/published

URLs for RTMP — important points

Important The URL for publishing over RTMP requires your attention.

When the RTMP stream name is short, a third-party software may not be able to recognize which part of the URL is the name of the application, and which part is the name of the stream.

Flussonic Media Server automatically omits rtmp from the application name. If the software requires you to specify some kind of application name, you should use the name static. For example:

  • server URL: rtmp://flussonic-ip/static
  • stream name: published

If you use a composite stream name, like client15/published1, you can specify:

  • server URL: rtmp://flussonic-ip/client15
  • stream name: published1

With this configuration, you will be able to use all of the stream's settings and options.

Configuring publishing to a static stream via the web UI

You can add the 'url publish://' data source to a stream via the web interface.

To create a static stream with a published source:

  1. In the admin interface, create a stream: Media > Stream > add.
  2. Enter a Stream name.
  3. Specify publish:// as the Source URL. Alternatively, save the stream, go to the Input tab, and click accept under Published input.
  4. Click create to save the stream. OBS Studio

To specify additional options for a published source, click options:

publication options

Removing a published source from a stream

To remove a published source from stream settings:

  1. Click the stream in Media and go to the Input tab.
  2. Under Published input, click ignore. Alternatively, click remove next to the URL entry that contains the publish:// source. Media Server
  3. Click save to apply the settings.

Publishing with a dynamic name

Why use dynamic names and publishing locations?

You might want to use dynamic names for published streams if one or more of the following is true:

  • Your publications last for a limited period of time (unlike a 24/7 TV channel broadcast).
  • You manage a lot of publications, and it is too much work to create a separate stream for each of them.
  • You do not know the names of the incoming streams in advance. For example, you are dealing with a third-party application - like a web chat - that generates a new unique identifier for each stream that it publishes to Flussonic.

Flussonic solves these problems by allowing you to create a publishing location (publication prefix) where you can specify common settings for multiple streams.

A dynamic name means that the full name of a stream is formed from a pre-configured publication prefix and the name defined in an external app.

If the name of a published stream is not known beforehand, or if you expect many published streams, you should set up a publication prefix:

live chats {
}

Here, chats is the publication prefix. All streams published under the chats prefix will have settings that you specify in the live directive.

All stream settings are described in Stream settings.

Publishing via WebRTC to live locations

You can use the following transcoder settings with live WebRTC published streams:

  • output_audio=(aac|aac_opus|opus). Specifies audio transcoding options. You can set up the resulting stream for playback in AC, AAC+Opus, or Opus. By default, aac_opus is used – input streams, which usually have audio in Opus, receive also a track in AAC on the output.

  • prefer_codec=(h264|vp8|vp9). Offers to choose one of the listed video codecs at the start of the publication. The default codec is H264. Video encoded in VP8 and VP9 is currently transferred without transcoding, its support will be added in the future.

Example:

live chats {
    output_audio aac;
    prefer_codec vp8;
}

URLs for publishing via different protocols

In case of publishing with a dynamic name, you will need to publish streams under names with a prefix, for example:

  • rtsp://flussonic-ip/chats/tempname
  • http://flussonic-ip/chats/tempname/mpegts
  • rtmp://flussonic-ip/chats/tempname

The part of the name that goes after chats is defined in the client app. Flussonic Media Server does not "know" the stream name in advance.

Setting up publishing with a dynamic name via the web interface

You can use the web interface to create a publishing location (a prefix for publication under dynamic names.)

  1. Go to the Media tab: publish-media
  2. Find the Publish locations section and click add: Publishing video to the server
  3. Enter a prefix and save the configuration. The prefix must be unique and may contain Latin characters, digits, underscores (_), and slashes (/). It is strongly recommended that you avoid using other characters, because client applications might not allow their use. (Flussonic itself has no restrictions on naming prefixes.) Publishing video to the server

Publishing via RTMP

There are several difficulties that come with publishing via RTMP. The RTMP protocol does not support proper URLs. When connecting to a server using RTMP, the client must specify both the server name and the application name, and then specify the name of the stream to be played or published to.

Usually, the shortened version is used where the application name and the stream name are concatenated. For example, the pair rtmp://rtmp.myhosting.com/chats and chat-15 is turned into rtmp://rtmp.myhosting.com/chats/chat-15.

The standard practice when interpreting the RTMP pseudo-URL rtmp://rtmp.myhosting.com/chats/chat-15 is to truncate the path after the server name at the first slash symbol and use the first part of the string as the application name.

This approach is being used only for RTMP and Flussonic Media Server does not support this concept.

Instead Flussonic uses the following logic when publishing via RTMP:

  1. The server concatenates the application name with the path being published. Thus, the pairs rtmp://rtmp.myhosting.com/chats/my, chat-15 and rtmp://rtmp.myhosting.com/chats, my/chat-15 produce the published stream name chats/my/chat-15
  2. The program searches for the first publishing prefix this name contains. In our example, that would be the prefix chats
  3. Then, all authorization interfaces and the like use the complete stream name: chats/my/chat-15.

To test that publishing over RTMP works, you can use ffmpeg:

ffmpeg -re -i /opt/flussonic/priv/bunny.mp4 -vcodec copy -acodec copy -f flv rtmp://localhost/chats/my/chat-15

This command should cause a new stream to appear in the web interface:

publish-stream

Publishing via RTSP

Some clients can publish video over RTSP.

Flussonic Media Server supports automatic selection between UDP and TCP transport and will receive the stream using the protocol selected by the client.

The stream name must be complete: chats/my/chat-15

ffmpeg -re -i /opt/flussonic/priv/bunny.mp4 -vcodec copy -acodec copy -f rtsp rtsp://localhost/chats/my/chat-15

Publishing via MPEG-TS

When transcoding a stream using ffmpeg, it is possible to publish video over HTTP. Video can be published with mpegts added at the end of the URL:

ffmpeg -re -i /opt/flussonic/priv/bunny.mp4 -vcodec copy -vbsf h264_mp4toannexb -acodec copy -f mpegts http://localhost:8080/chats/my/chat-15/mpegts

Authorization for stream publishing

Flussonic Media Server server can verify a password when publishing a stream. Enter the password in the configuration file as follows:

  live chats {
    password mypass;
  }
  stream published {
    password secure;
    url publish://;
  }

RTMP

To publish a password-protected RTMP stream, use the following example: rtmp application rtmp://192.168.2.3/live and stream name mystream?password=mypass

HTTP MPEG-TS

To publish an HTTP MPEG-TS stream, you can enter the data as follows:

    http://192.168.2.3:8080/s1/mpegts?password=secure
    ffmpeg -re -i video.mp4 -vcodec copy -acodec copy -f flv rtmp://192.168.2.3/live/mystream?password=mypass
    ffmpeg -re -i video.mp4 -vcodec copy -bsf h264_mp4toannexb -acodec copy -f mpegts http://192.168.2.3:8080/s1?password=secure

Extended publish validation

Flussonic Media Server allows you to configure an HTTP handler or a user script handler that will look at additional information about the publisher before accepting or rejecting the published stream.

live chats1 {
  on_publish http://mock:5000/publish-check;
}
live chats2 {
  on_publish /etc/flussonic/publish_check.lua;
}

Flussoniс will send an HTTP POST request with a JSON body to a HTTP handler, or pass a req object to a Lua script.

Both the JSON request and the req object have the following fields:

  • name — the name of the published stream (like chats/15), including the publish prefix.
  • location — the publish prefix.
  • proto — the publishing protocol (rtmp, rtsp, mpegts)
  • args — parameters extracted from a request's query string. For RTMP, a query string is extracted from the name of published stream.
  • ip — the IP address of the publisher.

An HTTP handler must return a response with status code 200 to start publication. The response may contain an x-record header with a file path.

A Lua script handler must return a pair of true and an object to start publication. The object may contain a record_path field with a file path.

Important! If a handler specifies a filename, Flussonic Media Server will start writing an flv file. It will keep writing to the file for as long as the stream is active, and there is disk space left. Flussonic will not check whether it is going to overwrite an existing file with the same name.

DVR archives and dynamic names of streams

You can configure the DVR archive for a live prefix:

live recorded {
  dvr /storage 3d 500G;
}

In this case, the published video will be recorded, and will be available even if the publication is terminated.

When the client stops publishing the video, the stream will disappear after some time, and Flussonic Media Server will keep a very little information about it. Information about this stream will be stored in the index of the archive, and Flussonic Media Server will not lose the files on the disk.

If configured, the system of purging of the archive will delete published streams according to the schedule.

Re-publishing

When you use publication prefixes, you can't use push with UDP:

live pushed {
  udp 239.0.0.1:1234;
}

Or rather you can, but in this case all streams will be sent to the same multicast group which will lead to complications.

Instead, you should use push with a template (%s) for re-publishing a stream via other protocols:

live pushed {
  push rtmp://cdn-server/client43/%s;
}

With this configuration, to republish the stream pushed/mystream, Flussonic will use the URL rtmp://cdn-server/client43/pushed/mystream.

Configuring adaptive bitrate (ABR) for WebRTC publications

When a client device is publishing from a browser to Flussonic, Flussonic can control the browser from which the publication is carried out so that the browser adjusts the bitrate of the publication to the bandwidth of the channel. This prevents packet loss when the bandwidth of the Internet connection is insufficient. If you reduce the channel width, the client must reduce the publication bitrate, if you expand the channel, the client must increase the publication bitrate.

To change ABR settings via the Flussonic UI:

To specify additional options for a published source, click options next to the source URL. Adaptive bitrate settings are under WebRTC > ABR:

publication options

The following settings in the file correspond to these settings in the UI:

stream published_stream_name {
  url publish:// abr_loss_lower=2 abr_loss_upper=10 abr_mode=1 abr_stepdown=50 frames_timeout=1 max_bitrate=2200 min_bitrate=500 output_audio=aac priority=0 source_timeout=5;
}

Options for adjusting the bitrate from the browser-publisher to the channel width up to Flussonic (ABR)

  • max_bitrate

    The maximum allowed bitrate.

  • min_bitrate

    The minimum bitrate threshold, by default 100kbit/s.

  • abr_mode (percent)

    The algorithm for determining the need to change the bitrate of the published stream and for calculating the target bitrate.

    abr_mode=0 - Flussonic's algorithm (similar to the one implemented in Chromium). This mode takes into account the packet losses, target bitrate, browser bitrate and the number of auto-adjustment cycles.

    abr_mode=1 - In this mode only packet losses and target bitrate are taken into account.

  • abr_stepup (percent)

    Increment step for raising the bitrate to the maximum, by default is 30%. If the loss is < abr_loss_lower, Flussonic makes the publisher to step up from the current bitrate to the maximum one with the rate of abr_stepup percent.

  • abr_stepdown (percent)

    The step of reducing the bitrate to the minimum, in percent. If packet losses > abr_loss_upper - Flussonic makes the publisher to reduce the current bitrate in steps with the maximum rate of abr_stepdown percent.

  • abr_loss_lower (percent)

    The lower limit of packet loss. When it is reached, Flussonic raises the bitrate. That is, if packet loss is less than abr_loss_lower, Flussonic makes the publisher to step up from the current bitrate to the maximum one with the rate of abr_stepup percent.

  • abr_loss_upper (percent)

    The upper limit of packet loss. When it is reached, Flussonic raises the bitrate. That is, if packet loss is greater than abr_loss_upper - Flussonic makes the publisher to reduce the current bitrate in steps with the maximum rate of abr_stepdown percent.

  • abr_cycles

    The number of cycles of bitrate auto-adjustment.

    Flussonic recommends the browser bitrate within the min_bitrate -max_bitrate range, depending on the presence and amount of packet losses during publication. Flussonic recommends lowering the bitrate when the amount of losses is more than abr_loss_upper and increasing it when the amount of losses is less thanabr_loss_lower. Decreasing and increasing occurs in steps of size abr_stepdown andabr_stepup, respectively. After the specified number of auto-adjustment cycles (abr_cycles) has passed, Flussonic considers the bitrate to be optimal and no longer analyzes it. By default, abr_cycles=5. If abr_cycles=0, the adjustment process takes place all the time while the publication lasts.

    Also, Flussonic calculates the actual maximum bitrate. It remembers the bitrate values at which the losses grew up to abr_loss_upper and considers their average value over the past number of cycles to be the new maximum bitrate value (actual).