Skip to content

Live Streaming

Flussonic Media Server can retransmit streaming video into multiple output formats on the fly with just-in-time packaging. For example, you can ingest an MPEG-TS stream, deliver it simultaneously to thousands of subscribers in DASH or HLS format, and at the same time publish the stream via RTMP on YouTube.

Flussonic Media Server supports three types of streams:

  • static — streams that are being broadcasted all the time.
  • ondemand — streams that are requested by users (on demand).
  • live — user-published streams. See Publishing for details.


Static streams

Static streams are launched upon a start of the server. Flussonic continuously monitors static streams.

If for some reason (transcoder went off, antenna broke down) a data source goes down, Flussonic Media Server will constantly keep trying tо reconnect to the stream until success or shutdown.

Usually, IPTV channels or IP camera feeds are being configured as a static stream.

Flussonic Media Server supports many types of data sources, which must be configured with URLs.

The format of a stream definition in the /etc/flussonic/flussonic.conf file is:

stream example_stream {
  input udp://;

In this example:

  • example is the name that must be used to request the stream from Flussonic Media Server.
  • udp:// is the the data source URL.

Important. The name of a stream should contain only Latin characters, digits, dots (.), dashes (-), and underscores (_). If the name contains any other characters, DVR and live streams might work incorrectly.

To add a stream via the web interface:

Go to the Media tab and click Add next to Streams.

Flussonic add stream

Then enter the name of the stream and the data source URL. Click Create and Flussonic will add the stream to the list.

Note. By default, new streams are Static. To change the stream type to On demand, click Static next to the stream name.

After the stream is added, you can go to the stream settings page and check the ingest status:


On-demand streams

If the stream is not needed all the time but only upon user's request, you can configure Flussonic Media Server to turn it off when it is not being used and turn it back on when it is requested.

To specify this kind of behavior, change the stream type to ondemand:

ondemand ipcam {
  input rtsp://localhost:554/source;

Flussonic ondemand


If Media Server ingests the ondemand source stream using RTMP, RTSP, or HTTP MPEG-TS protocols, there will be some complications with outputting HLS streams. This is because those streaming protocols require 10-30 second buffering. The player will not start playback until its buffer is filled, so the first user who initiated playback would need to wait until the system is ready. The only data source that doesn't have this problem is another Flussonic server with HLS protocol. Flussonic Media Server uses its own extensions that allows for instant playback on iPhone.

You can specify the stream's lifetime after a client has disconnected:

ondemand ipcam1 {
  input rtsp://localhost:554/source;
  retry_limit 10;
  clients_timeout 20;

The config line above has the following meaning: make no more than 10 attempts to reconnect with a data source if the connection is lost; when the last client leaves, run keep fetching the stream for no longer than 20 seconds.

Streams playback

How to playback streams, learn in Video output.

Stream screenshots in JPEG

Flussonic Media Server can generate JPEG thumbnails of a streaming video. To use this feature, add the thumbnails option in the stream settings:

stream example {
  input fake://fake;

Alternatively, to reduce CPU usage, you can specify an URL where Flussonic Media Server can get JPEG thumbnails. Many IP cameras have a special URL with screenshots:

stream example {
  input rtsp://localhost:554/source;
  thumbnails url=http://examplehost:5000/snapshot;

You can find the screenshot URL in the documentation for you camera model.

Flussonic JPEG thumbnails

The latest screenshot of a stream is available at http://flussonic:80/example/preview.jpg

An MJPEG screenshot of a stream is available at http://flussonic:80/example/preview.mjpeg

See also:

Substituting a stream with a file

If a stream becomes unavailable, Flussonic can substitute it with a fallback video from a video file that you specify using backup <VOD location>. This works for any live streams, including published ones.

stream example {
  input tshttp://;
  backup vod/bunny.mp4;

You need to specify the path to the fallback file relative to the VOD-location, for example vod/backup.mp4, where vod is the unique name of our VOD location. Do not use absolute paths for video files.


If the original stream has no audio (for example, a stream from an IP camera), the substitute file must have no audio as well.

By default, the fallback file is not recorded to the archive and is not transcoded. However, you can configure it.

To learn more about different ways of using files as stream failover sources, see the section Source Failover.

How 'backup' is different from 'input file://'

Unlike source switching with the source input file://<VOD location>, when a fallback file is used, Flussonic technically does not switch to another source. This is especially useful for published streams to prevent numerous closings of a socket with a publishing client.

The fallback file specified in backup <VOD location> is not transcoded and not written to DVR, unless you configure otherwise. The file source input file://<VOD location> is always written to DVR.

When use backup instead of input file://:

  • In case of poor connection with the client that publishes video, Flussonic continues to receive frames without interrupting the connection with the client. This allows the client to continue the publishing session without having to start it over each time the source was switched. When the published stream disappears, viewers see the fallback file and understand that the broadcast is not over yet.

  • When all sources are unstable and Flussonic switches between them too often, it is better to show a fallback file. If you use a file as one of the sources, viewers will see any video only after timeouts pass for each of the troubled sources.

  • If you write the main stream to DVR and do not want to write the file too in order to prevent the file from appearing in the archived video.

  • Using options like timeout for the main stream and the fallback file, you can manage which source to show during a publication session.


Sometimes names of streams on a remote server are not known in advance, so you want to dynamically rewrite the source (input) URL based on the requested stream name. To do that, use the following template feature:

template nsk {
  prefix nsk;
  input rtsp://streamer:555/%s;
template ams {
  prefix ams;
  input hls://streamer:8081/%s/index.m3u8;

The template directive uses the %s pattern to replace a substring in the input URL. It means that the substring that succeeds the prefix specified in the template's prefix replaces the %s pattern in the source URL, i. e. a substring that goes prior to the prefix, including the prefix itself, is omitted and what is left applies to the input URL. Flussonic receives a request for one URL but requests a different URL from the streamer without a redirect.

For instance, if a client requests a stream over the following URL: http://FLUSSONIC-IP/ams/ort/index.m3u8, Flussonic delivers the stream from the source URL: hls://streamer:8081/ort/index.m3u8, so the http://FLUSSONIC-IP/ams/ substring is removed, and the ort/index.m3u8 substring is used in the source URL input.

How do I use it?

For instance, to broadcast the channels to various areas from different servers, i. e. one server for every area. It provides you with a quick and efficient way of content delivery to the viewer.

Recording video streams (DVR)

Flussonic Media Server has a built-in state-of-the-art stream recording system.

The stream archiver can record video, provide access to a particular video interval, export parts of the archive as MP4 files, clean up old archive files, and maintain ample free space on the storage disk.

To turn on the archiver, specify the dvr option in a stream settings:

stream foxlive {
  input tshttp://trancoder-5:9000/;
  dvr /storage 90% 5d;

For details, see archive management.

Time zone adjustment (Timeshift)

Flussonic Media Server can play the archive record of a stream with a fixed delay.


Flussonic Media Server maintains the exact delay, so if for some reason the archive has gaps, end users will be getting no video for the duration of a gap.

The timeshift feature has its own data source schema — timeshift://:

stream channel {
  input fake://fake;
  dvr /storage 90% 5d;
stream channel-2h {
  input timeshift://channel/7200;

The delay is specified in seconds.

Flussonic timeshift

Stream delivery over UDP multicast

Flussonic Media Server can rebroadcast a stream from a data source over the local network.

Flussonic Media Server demonstrates next-to-ideal jitter values when streaming multicast UDP over the network.

stream example_stream {
  input tshttp://localhost:80/origin/mpegts;
  push udp://;

Stream settings for IP surveillance cameras

It is possible to configure Flussonic Media Server to request a stream from camera via UDP only. This might be useful when dealing with cameras that have issues with streaming over TCP.

stream cam1 {
  input rtsp://localhost:553/bunny.mp4;
  rtp udp;


The HEVC (H.265) video codec is supported only in Chrome (version 107 and higher), Microsoft Edge (version 16 and higher), and Safari (version 11 and higher) on desktops, and in Chrome (version 107 and higher) and Safari for iOS (version 11.0 and higher) on mobile devices. All other browsers cannot play H.265 video streams. More on this in Playing H265.

If there is no need to retrieve audio from a camera (for example if the audio is encoded in G.726), you can configure Flussonic Media Server to ingest only one track. The number of the track must be specified in the stream settings:

stream cam1 {
  input rtsp://localhost:554/origin tracks=1;

In order to transcode an audio stream from G.711a or G.711u into the AAC codec, use the protocol rtsp2:

stream cam1 {
  input rtsp2://localhost:554/origin;

Turning on audio-only HLS

When approving apps for publishing in AppStore, Apple may require the stream to have an audio-only version. To satisfy this requirement, add add_audio_only directive to the configuration:

stream cam1 {
  input fake://fake;

With this directive, if the stream contains both audio and video, Flussonic will generate multibitrate playlist with two profiles - one with audio only and another with audio and video tracks.

Capturing stream from another Flussonic Media Server

The details of transferring video between Flussonic Media Server servers are discussed in Flussonic video stream clusterization.

DRM in live streaming

The details are discussed in the DRM article.

Silence detection

Flussonic can detect low sound level (no sound) in sources of input streams and notify about it. See Silence detection for details.

Stream and group settings

These settings are used in the directives stream, ondemand, and live. We call them options.


on_play http://backend/; Enables authorization for a stream. See more in the authorization section.


domains *; Specifying the domains, within which playing this video is allowed. This does not work for those clients that do not pass the value of Referer. To work correcty in the WEB the flussonic domain must present in the list (the domain of the embed.


allowed_countries CA US UK; The list of two-character codes of countries where the access is allowed (for code reference see the MaxMind database).


input tshttp://transcoder:port/; URL of the data source. It is possible to list several URLs to enable trying the first available data source.


If a UDP source is used, the configuration file must contain this particular UDP address only once. If multiple streams use the same UDP address, chances are it will not work.


url_prefix prefix

for example

When using HLS or DASH protocol, the addresses of individual segments and playlists within the variant playlist will start with the specified prefix. This option may be used not only as part of an individual stream's settings but also in the global portion of the config file. If the option is specified globally, it will be applied to all streams on the server.


dvr /storage 1d 50% schedule=8:00-16:00;

dvr @my_raid 1d 50% schedule=8:00-16:00;

Enables writing a stream to a DVR archive. The first example indicates that Flussonic Media Server should store the archive in the /storage/streamname directory. The second example configures the server to store the archive in the disk array @my_raid. Flussonic will clean up that directory either once a day or when the disk gets 50% full. To set up storage time, use days or hours, for example, 20h. The parameter `schedule` allows you to set a schedule of recording, in the form of intervals. The time is specified in UTC in hours and optionally with minutes; the interval can last after midnight: 22-1:30. A schedule can contain multiple intervals, separated by a comma: 8:00-16:00,22-1:30.


dvr_offline /storage 1d 50%; With this option specified, Flussonic will not write the stream to the archive until you turn on the archiving of this stream explicitly via the API. This option is used in place of the dvr option.

push udp

push udp://; This option makes Flussonic Media Server to send the stream via MPEG-TS over UDP.

To set MULTICAST_TTL parameter on the UDP socket, use the following syntax: push udp://;.

To set constant bitrate (CBR) use: push udp://;, where 2000 is the bitrate in kbit/sec.

The option multicast_loopallows you to ingest the stream sent to UDP multicast back on the sending Flussonic host.

If the transcoder was enabled for this stream, Flussonic first transcodes the stream and then pushes it.


thumbnails; Turns on generation of the stream preview JPEG thumbnails.


retry_limit 10; Sets the number of times Flussonic Media Server will try to connect to the data sources before closing a non-static stream.


clients_timeout 10; Sets the time period (in seconds) for which Flussonic Media Server will keep serving a non-static stream after the client's last request.


source_timeout 10; Specifies the period of time, in seconds, for which Flussonic Media Server waits for new frames to come from the data source. When this time passes, Flussonic attempts to reconnect to the data source. Default source_timeout is 60 seconds.


frames_timeout 3; Specifies the period of time, in seconds, for which Flussonic Media Server waits for new frames to come from the data source before it generates the event frames_timed_out. This period of time must be smaller than in source_timeout. The event frames_timed_out informs you that the source might soon be lost. If frames come again from this source, before source_timeout has passed, Flussonic issues the frames_restored event.


password secret; The password that will be passed via query string (http or rtmp) for publication in a stream or group.

push rtmp

push rtmp://DESTINATION_SERVER/STREAM_NAME; With this option Flussonic will publish the stream to another server via the RTMP protocol. If the transcoder was enabled for this stream, Flussonic first transcodes the stream and then pushes it.


backup vod/blank.mp4; Setting this option for the stream will launch the specified vod vod/blank.mp4 in case the video from the data source becomes unavailable. You can manage the backup file with additional options additional options.

input publish://

input publish://; This option is used for publishing video into the stream. This option is not applicable for stream groups.


on_publish http://host/publish.php;

on_publish /etc/flussonic/publish.lua;
Enables callback script or http push event when video is being published to this stream or group. HTTP push event could contains information about the stream name, publisher's IP, etc. In response it is possible to allow or deny the publication: the HTTP backend must return 200 OK or 403 Forbidden; the .lua script must return {true, {}} or {false, {}.


max_sessions 1000; Sets the limit on the number of sessions for the stream.


rtp udp; Turns on mandatory use of UDP for communicating with RTSP cameras.


add_audio_only; Adds to the HLS playlist a link to an audio-only stream. This is needed to validate an app in Apple devices.

prepush off

prepush 0; Turns off the live-stream_settings_prepush">prepush feature (used for a quicker stream start). This option is useful for broadcasting real-time streams.


prepush 10; Enables a buffer of a specified duration, in seconds. If the client's connection to the server is interrupted or slowed down, it will keep playing video from the buffer. This allows the player to start faster, but with a delay relative to the source.


max_bitrate 1000; Sets the bitrate limit for the stream that is being published.


logo path=flu/embed-logo.png height=100 width=100 left=0 top=0; Add logo at playback. This logo will not be displayed on mobile devices and in the DVR player. To add logo to the video use transcoder.
path (required) — path relative to wwwroot directory.
height, width — logo image size in px. If ony only one of these parameters is present then the other is scaled proportionally. Omit these parameters to display logo in the original size.
left, top, right, bottom — logo image location specified by offset in px. For example, right bottom corner: right=0, bottom=0. Don't use left and right, top and bottom parameters together.


mpegts_pids pmt=4095 sdt=0x12 v1=211 v2=212 a0=220 t0=16#fb; This parameter sets PIDs values for outgoing MPEG-TS streams. It is possible to set PID values for PMT, STD and video and audio tracks. Tracks are numbered starting from one. The code a1=123 sets a PID value for the first audio track. It is possible to set base index for the tracks of certain type using the 0 (zero) index. Example: `t0=100` sets PID=101 for the first track, 102 for the second, and so on. Numbers can be given in decimal form (by default) or in hexadecimal with 16# prefix.

There is one more way to list PIDs.


meta program_id "10"; This parameter sets the program identifier in the output MPEG-TS stream.


provider Flussonic; This parameter sets the program provider title in the output MPEG-TS stream.


segment_duration 4; Specifies the duration of a segment for HLS streams in seconds. For some incoming streams Flussonic will not apply the specified segment duration. This depends on a stream's GOP duration in seconds. A segment duration must be divisible by GOP, because GOP structure cannot be split into smaller parts. For example, for a stream with 4-second GOPs, possible segment duration is 4 seconds, 8 seconds, 12 seconds, and so on. Otherwise, Flussonic will create segments equal to each GOP in a stream.


segment_count 4; Specifies the number of segments in HLS playlists.


group sport; Used only on a source server to define the names of TV channel groups where the stream is included.


disabled; Disables the stream.