Bugfix release 3.9.2 out now!

Bugfix 3.9.2 Release notes

Hey Everyone, we've got a surprise 3.9.2 release as there were a few issues that we found quite necessary to patch and a few extras that we had ready to go that didn't quite pass testing before 3.9.1. The changelog will is also available.

Proxied IP addresses are now correctly passed again (regression introduced in 3.9).

Proxied IP addresses always passed on the address of the proxy instead of the proxied address. This has been fixed.

API port will now also consider X-Forwarded-For to consider traffic non-local.

This makes a reverse proxy for the API port require a login as well. Instead of only X-Real-IP. This is actually something we thought we had implemented, but apparently only did so for the non-API ports.

RIST profile setting can now be set.

RIST ignored the simple profile, this can now be set again correctly.

Wildcard Support for certificates, fixed default certificate fallback

Wildcard certificates weren't properly recognized and wouldn't match the wildcard portion. If no match can be found the fallback is to go to the first certificate that is set up.

An Alpine-based Dockerfile is now part of the main repository.

This can be used to create a MistServer docker image from the source code, making it easier for our users to create their own docker images.
Though the usage is likely clear if you're familiar with Docker expect the documentation to be updated next week at the latest.

The HTML output now supports several new URL parameters.

Fillspace is now the default, which makes the player take up the maximum space it is allowed to.
dev and muted no longer need a value, they only need to be passed.
The documentation has been updated to reflect the changes.

Bugfix 3.9.1 Release out now

Bugfix 3.9.1 Release notes

Hey everyone, bugfix release 3.9.1 is now available. We've found a few things we wanted changed, but not enough to consider as a full release. If you want to know more we recommend looking at the full changelog.

Downloads can be found here.

Noteworthy Changes

Logout API call

There's a new API call to log out, which forces a clear of the authentication state of an API connection. This is necessary for browsers that can re-use previous connections even after log out

Track-ID based track selector

You can now use the track id (or pid) of a track within a stream. This works for video, audio and metadata. It can be done by adding an i# as selection. For example selecting the video track 256 you can use ?video=i256 for more information see track selectors.

Improvement EFPS / EFPKS

The effective frames per second and effective frames per kilo second have been improved to be caucluated over time rather than a single measure. This makes them far more accurate, though it can take up to 10 seconds before it's done measuring.

Metadata improvement

Stream health information now includes the track index and the (effective) frames per second.

Shared memory statistics in capabilities API call

Shared memory statistics are now available in the capabilities API call. Visible on the server stats page

Simulate live option is removed from the interface

This was an option that was never meant to be user-facing rather set automatically depending on the input used. Most notable playlist inputs, the option has been removed as it can cause unexpected/wrong behaviour if set wrongly.

Bugfix: JWK memory leak

There was a memory leak if JWKs were loaded from an URL. This has been patched.

Bugfix: DTSC memory leak

DTSC output could cause a minor memory leak when connections did not have enough bandwidth.

Bugfix: UDP API no longer respons to requests coming from its own listening port. This could sometimes happen and caused 100% CPU usage due to a loop.

Push statistics no longer accidentally pick up non-push SRT connections

Several other bugfixes and improvements.

Stable release 3.9

Stable 3.9 Release notes

Hey everyone, we've just released our 3.9 update! It brings some exciting new features such as Enhanced RTMP (E-RTMP), native JWT support, and some very basic SCTE35 support for TS-based outputs. All in all there's quite a few changes, so you might want to go through the full changelog.

Downloads can be found here.

Noteworthy Changes

Enhanced RTMP (E-RTMP) following 2.0 Specification

The full Enhanced RTMP V2 specification is now supported by MistServer allowing all features to be used. The upgrade happens automatically when needed and connections will automatically upgrade should the other side be capable of it as well. Both input and output support all aspects of the specification.

JWT support (JSON Web Tokens)

JWT support is now included in MistServer, which is a great method to bind multiple applications/platforms together with a unified authorization method. Currently you can only use JWTs to stream into MistServer or pull streams from MistServer. There are plans to add support for accessing the MistServer API (and web interface) as well. You can find the JWT developer documentation in our online documentation. A basic integration example can be found here.

WHEP player now included in meta-player

WHEP WebRTC output can now be played directly in the MistServer Meta-player. Of course this is in addition to WebRTC using Websockets for signaling: both methods will be supported going forward.

Meta-player keyboard controls

We've added keyboard controls to the meta-player. Various playback features such as pause, play, speed up, speed down can be done through the keyboard. Controls will be enabled by default unless "keyControls":false is set.

Stream processes automatically masked from other processes

Tracks created by stream processes will now default to being masked for other processes. This was changed since most users do not expect to re-use the result of, say, a transcode for other processes. Keep this in mind should you create a track that is meant as a source for another process, as you will have to change your configuration to oppose the new default.

RTMP improvements

RTMP push outputs can now gracefully or gracelessly disconnect. A graceless disconnect is not informing the other side that the stream is about to end, which can trigger recovery logic on the other side. Graceful disconnect means informing the other side of the stream end as before.
When pushing RTMP out you can now override all aspects of the target URL by the host, application and stream URL parameters. This allows the usage of non-default URLs that might not be parsed correctly otherwise.
You can also use the keepts parameter to keep the stream's internal timestamps unchanged instead of normalizing them to start from 0 when pushing out.

Jitter / Latency improvements

We've made our jitter/latency handler more aggressive in trying to lower the latency as much as possible if the connection stability allows for it.

Very basic SCTE-35 support for TS-based outputs

Basic support for creating SCTE-35 markers in TS-based outputs from JSON instructions. Sending in a simple JSON post towards MistServers meta channels will create a SCTE-35 marker which will be inserted in TS based outputs.

You can add SCTE-35 markers by sending in a JSON post towards:

http://mistserver:httpport/streamname.json?meta=SCTE35

With Content-Type set to application/json and a POST body of {"splice_out":DURATION} where duration is in milliseconds. The splice-out point will be inserted immediately.

PLAY_REWRITE trigger

A new trigger that lets you change which stream is loaded for any request, letting you overwrite the URL format arbitrarily in the same way that PUSH_REWRITE already lets you do for incoming pushes.

You can now use ! as negated track selector

Using ! at the start of a track selector is now shorthand for all,!: selecting all tracks that do not match with the given selector. For example video=!h264 would match every video track except for h264-encoded tracks.

MistUtilHealth utility intended for automated Docker container health checks

A new MistUtil has been added: MistUtilHealth. It behaves in a way compatible with Docker container health checks, and will soon be automatically used for our standard Docker containers.

Effective frames per second track property

Effective frames per second property has been added that contains the measured FPS rather than the one signaled in the track metadata. The new property can be found in all places the "regular" frames per second property can also be found.

Pushing improvements

Live push outputs now have a latency measurement, which is how close to the theoretically perfect live point the push is currently sending data from. (This makes it effectively an "internal" latency measurement, rather than "end to end" latency measurement.)
A waitkey URL parameter can now be used to wait for the next keyframe before starting a push output, to minimize latency even further.

Live input improvements

Live streams can now configure a "max track idle time", which is how long a track can be idle before it is dropped. Before, this time was synchronized to the "DVR window" setting. They can be independently configured now.

Updated to mbedtls 3.6.5

We've upgraded our own builds and the meson subproject (for compiling yourself) to the latest mbedtls 3.6.5 stable release.

Fixed support for multiple SSL certificates/keys

You can set multiple certificates/keys for a single MistOutHTTPS and/or MistOutRTMPS listening port now. This allows you to serve multiple domains from a single port, even if you need to use more than one certificate to serve them (automatic TLS SNI support).

Various other improvements and bugfixes

We've massively improved the RTMP input and output protocols and done various other changes that improve compatibility with other platforms as well as bugfixes. Full details in the changelog!

Stable release 3.8

Hey everyone, we've just released our 3.8 update! A long standing bug where ?stream=STREAMNAME got added to any HTTP output has been resolved. We've improved support for VP8/VP9/Opus and AV1 and made our UI much faster by making it eventbased rather than threaded.
We've also added methods to use tags and streamname(+wildcard) to make selections.

Downloads can be found here, and the full changelog can be found here.

Noteworthy Changes

HTTP-based URLs in triggers no longer contain ?stream=STREAMNAME

If one of your trigger handles expects this value to be in the URL you will need to update it accordingly. You can still parse the the unprocessed stream name from the URL yourself, or take the processed stream name from the rest of the trigger payload.

Auto-pushes can now be disabled based on specific tags or stream names used as inhibitor.

You can now specifically use #tags or streamname(+wildcard) to not activate the pushing logic

"inhibit":["live+example","#replicated"]

This would invalidate the live+example specifically and any live+wildcard stream with the #replicated tag for this push.

Live stream processes can be disabled through tags and triggers

Similar to the auto-push function you can also select specific streams/tags to not start processes on.

Triggers can now be disabled based on specific tags or stream names.

This one is slightly different compared to processes and pushes. This functionality is in the Applies to field, which is why we've also added a method to unselect a specific #tag or streamname by adding ! in front of it: !#tag and !streamname(+wildcard). This selection will be ran from left to right and can re-activate certain streams to work again. For example first removing all live+wildcard through !live+ then specifically adding live+example to only work for live+example.

DTSC-pulled streams are now automatically tagged with the #replicated tag

This makes it easier to both see which streams are coming from another Mistserver as source, but also to use them for process/pushing/recording logic.

VP8/VP9/Opus support in MP4 input/output

For VoDs that contained any of these codecs you will need to delete the matching DTSH file in order to see the tracks with these new codecs. The old DTSH files are usable, however only the "old" tracks will be used.

AV1 support in RTP (WebRTC, RTSP, SDP) and TS-based protocols (where available).

Note that the AV1 in TS specification is not finalized yet and our output implementation is technically incompatible (though the input implementation is fully compatible with both spec-compliant and our own output). Problems and incompatibilities might occur accordingly, but we do plan to be fully specification compliant in the near future.

3.7

Hey everyone, we've just released our 3.7 update! Some good improvements all around, new feature "stream keys" and WHIP push to other servers!

Downloads can be found here, and the full changelog can be found here.

Noteworthy Changes

New feature: Stream keys

Stream keys adds native support for the most common method of pushing live streams into MistServer. It allows you to set or generate a token for stream names and thus prevent the need for further customization or integration if this is your intended usage.

For more information please look at the documentation. Specifically the stream key documentation.

New feature: WHIP push to other servers

You can now use WebRTC WHIP as a push option. You can do so through the following targets:

• http: whip://example.com/publish
• https: whips://example.com/publish

Should an authentication bearer token be needed you can add this through a parameter: ?auth=TOKEN

• http: whip://example.com/publish?auth=TOKEN
• https: whips://example.com/publish?auth=TOKEN

We have verified this implementation is cross-compatible with Cloudflare, Twitch and Pion (among others).

New feature: HLS push to Youtube

MistServer now supports HLS push to Youtube specifically. When MistServer recognizes a Youtube puhs url it'll automatically adapt the push output to work. Links such as https://a.upload.youtube.com/http_upload_hls?cid=TOKEN&copy=0&file= that come from Youtube Studio can now be used directly as a push target without further configuration needed.

New feature: Shell-style string quoting and escaping

This has been a limitation for quite a while. MistServer has various places where it would run commands that you specify. However it did not implement shell-style string quoting and escaping. This made it impossible (for example) to put spaces in arguments. You should now be able to copy most commands directly into the MistServer fields that run commands, without having to worry about changing the format to something non-standard.

Pull-styled streams function better

Sometimes pull styled streams (dtsc or srt pull) would run stuttery, several improvements have been made.

Local-only UDP API auto-retries connection

In some environments MistServer would lose contact to its UDP API. This is now able to reconnect properly.

3.6.1

Hey everyone, we've just released our 3.6 update! Most excitingly, we're finally back to having a unified version of MistServer across all platforms! It took quite a while, but we're finally there. A few final steps are still needed to bring them all to the same level of quality, but the biggest hurdle has been cleared.

Downloads can be found here, and the full changelog can be found here.

Noteworthy Changes

MistServer MacOS builds are available again

We've resumed our support for MacOS builds, and they are now available once more. They're full-featured and should be nearly identical to the Linux version, except they do not rely on shared memory. This results in only a slight performance hit.

MacOS builds may still complain on boot about having no available RAM, but that message can be safely ignored on that platform.

SRT working on Windows again

SRT should now work for both input and output once more. We managed to work around an issue caused by how Windows handles socket binding. The Windows builds use significantly more CPU time to serve SRT compared to the other platforms, but latency and the SRT protocol functionality itself are otherwise normal.

WebRTC on Windows

WebRTC on Windows is almost fully functional. Currently, it supports a single viewer simultaneously only due to a bug. We didn't want to delay the release for this, so it will be fixed in an upcoming patch.

WebRTC now has multi-path & single socket support

WebRTC now supports migrating connections between mobile and Wi-Fi networks and can operate from a single listening socket. This means you only need to expose one port for WebRTC instead of the entire ephemeral range. It's especially helpful when running MistServer in virtualized environments like Docker/Kubernetes, where forwarding large port ranges is tricky.
The setup required to get WebRTC to work has also been made much more simple, and the default configuration should work out of the box for most people.

Tag support improved

The active_streams call now returns stream tags, and you can now start pushes using tags by specifying #some_tag instead of a stream name. Automatic pushes will not start inactive streams, but a single push command will.

Push outputs will now report their current file target in the push status API

Previously, if dynamic variables were used, the API would only show the variable name instead of the final file target. You had to wait until the RECORDING_END trigger to know the result. Now, the push status API displays the final file target, and updates in real-time for segmented pushes.

The autopush API now supports JSON objects

The automatic push commands push_auto_add, push_auto_remove, and push_auto_list have been updated to use JSON objects. The old format remains available for backward compatibility. We'll be updating our documentation to reflect this change.

Push interface improvements

The push interface now allows you to enable/disable automatic pushes. When disabled, they won't trigger a new push under any condition, although existing pushes remain active. This makes managing pushes easier and gives more control.

Additional improvements include:

• Quick filtering for showing & stopping pushes
• Right-click options to edit, change, or stop pushes
• Clicking stream names to visit their preview pages
• Categorized push parameters and stream process options for a cleaner interface

Several other bugfixes and improvements

We've made various other minor changes to improve performance and user experience. We recommend checking the full changelog for the complete list.

3.4 stable release

Downloads can be found here and the full changelog can be found here.

Noteworthy changes

Windows builds!

We finally have releases of Windows builds for 3.X, starting with this version. The MistTray taskbar icon application has not yet been updated, so this release does not include it.
MistServer should now be ran using the MistServer Shell shortcut. This boots MistController in a shell window (which you can minimize to get it out of your way, or keep in view to see the logs in it).

There will still be a few differences in the Windows builds compared to Linux (most notably: WebRTC is not functional in the Windows build right now), but future updates will incrementally get Windows and Linux builds to parity again.

Web interface: New stream status page

The stream status page is a new tool in the web admin interface to more effectively analyse the current state of a stream. It also allows the following actions:

• Add tags
• Stopping viewers
• Stopping sessions
• Nuking stream (forceful reset of the stream)
• Look at the meta-data of the stream
• Easily identify the live point for each track and their respective sync/delay
• Creating/removing triggers
• Starting/stopping pushes
• Starting/stopping recordings

Raw pixel support

Usually you don't want to handle raw pixels in a modern media server, but in some cases it is unavoidable. MistServer now supports a new custom data structure optimized for raw video streams, and WebM/MKV-based inputs/outputs/processes have been updated to support raw pixels.

V4L2 Support (Linux only)

You can now use V4L2 to directly access webcams and other compatible devices with MistServer. Both raw pixel mode and (M)JPEG mode are supported.

RTMPS push input and pull output support

You can now set up RTMP with an SSL certificate to support incoming RTMPS connections. (Outgoing RTMPS connections have been supported for several versions already.)

WebRTC datatrack support

WebRTC (both WebSocket-based signalling and WHEP signalling) now have support for datatracks to receive JSON-based data track messages in sync with playback. Ingest support and trick play signalling over datatracks will follow at a later date.

New settings for processes

The restart behaviour can now be configured, allowing you to set how processes should respond to failure conditions. Stream processes can now be marked as inconsequential (meaning they will not impact when a stream is considered available).

Goodbye Cmake

Support for building MistServer using Cmake is now officially dropped. MistServer can only be build using Meson starting with this release.

MistServer 3.3 stable release

Stable release 3.3 is now available! Downloads can be found here and the full changelog can be found here.

Noteworthy changes

Improved (live) metadata track support

Metadata tracks (subtitles, JSON side band data, etc) are now properly supported end-to-end. There are new player APIs to request metadata tracks be given to a JavaScript function in-sync with playback. Metadata tracks can now be muxed to and demuxed from private data tracks in TS streams (including SRT and RIST) for effortless replication across systems.
A new "protocol" allows easy ingest of subtitles and JSON sideband data tracks over standard input or TCP socket.

Lower live streaming latency

We reworked the internals of our live streaming core code, lowering the minimum end-to-end latency between input and output from at least 1 frame on each track, to at least 1 frame on any track. Especially considering metadata tracks tend to be pretty sparse, this can make a massive difference in terms of latency - making it possible to achieve sub-second latency with less effort than ever before.

Clipping support

It is now possible to create short VoD clips on-demand from both VoD and live streams, in any/all supported output formats (including MP4, MKV, WebM, TS, FLV, etcetera).

Chromecast and video pop-out support

The built-in meta-player now supports playing back to and remote controlling Chromecast devices, as long as the player was loaded over HTTPS.
It is now also possible to use the video pop-out feature some browsers offer.

HLS DVR input now full-featured

It is now possible to use an ongoing HLS-format recording as source for a stream that then will support both live playback and seeking into the full DVR window of the recording. This also works with remote HLS streams, both VoD and live. For now, the HLS streams must use TS segments (fMP4/CMAF segment support will come in a future version).

Exit reasons

There is a set of new triggers and log messages that track why a process shut down / stopped, which greatly eases debugging and allows for more reliable tracking if an operation succeeded or failed.

3.2.2 bugfix release

Bugfix release 3.2.2 is now available! We've made several minor improvements that we deemed important enough to release, but did not warrant a full stable release. Downloads can be found here and the full changelog can be found here.

Changes

• Feature: Support for AV1 codec in MP4 input and output
• Bugfix: RTMP push output sent initial data with wrong timestamps, fixed
• Bugfix: The new UDP packet pacing code accidentally broke WebRTC ingest, fixed
• Bugfix: A rolling restart will no longer shut down a large percentage of segmented-protocol (e.g. HLS) viewers
• Bugfix: Fix local file path DTSH writing (no longer written to a folder called "file:")
• Bugfix: HTTP parser no longer chokes on reading an empty response body in non-chunked transfer modes
• Bugfix: Incoming TS-SRT pushes set their UTC offset correctly now
• Bugfix: Fixed track masking feature in MistProcMKVExec
• Bugfix: If an TS-SRT push comes in and does not immediately send media data, no longer fails to receive the push
• Bugfix: Packet jitter calculator no longer shared between inputs of single-process inputs (e.g. TS-SRT, TS-RIST)
• Bugfix: Fixed output session shutting down while waiting for a pushdelay setting to be applied
• Bugfix: SRT and RIST listening sockets no longer kill all connections when there is a viewer for a mid-shutdown stream connected over these protocols
• Bugfix: Removed dead code in Meta-Player RAW WebSocket wrapper

3.1 release

The 3.1 release is upon us! Downloads can be found here.

First of all: apologies, this build is still Linux-only for now. While a lot of progress has been made on fixing MacOS and Windows compatibility in the past few months, it's not quite ready yet and we felt it prudent to release some of the fixes and features this release brings earlier rather than later.

This release also marks a new style of release post. We now mention noteworthy features with some text explaining what the feature is and why you'd want to use it, and of course the full changelog is still available as per usual.

Noteworthy features

TS-Based LLHLS

What: Support for Apple's LLHLS (Low Latency HTTP Live Streaming) protocol was already included in 3.0. However, that support is CMAF-based (the latest industry standard). HLS also supports an older TS-based segmenting method - this feature adds support for LLHLS in that format as well.

Why: This feature is useful for those that want/need to support LLHLS but are bound to using MPEG2-TS-based playback for some arbitrary reason (e.g. legacy device support, weird TVs they are using, etc).

Forward Error Correction

What: This adds support for ProMPEG Forward Error Correction to our MPEG2-TS-based output over UDP. FEC adds an additional layer of error correction packets on top of the normal packets, to allow for recovering from packet loss using parity data.

Why: Sometimes a connection can only be made over UDP, but there is a small amount of packet loss you want to correct for. This makes that possible. If you're not stuck using UDP, there are likely better solutions than this one.

Input SDP from file / Push output SDP to file

What: RTSP, WebRTC and SIP (VoIP) all use RTP-based transport for their media data. Some users want to bypass those signalling protocols and instead work with raw RTP over the network. The SDP input and output allows you to preconfigure the "handshake" normally done over the signalling protocol, so you can directly send or receive media without needing to do that handshake.

Why: This is especially useful when broadcasting multicast over, say, a company intranet - since the "handshake" should only be done once but there may be potentially infinite receivers in the network. They can all open the SDP file to receive the data needed to connect. Until this feature became available, multicast over the local network was not possible using RTP and only using raw UDP (in an MPEG2-TS transport). The support for SDP is not only multicast usecases, though - that is just the most noteworthy one.

AAC file input support

What: The AAC codec is already supported in Mist, but separate .aac files are not (they need to be inside some other container, instead). This update adds support for plain .aac files as a VoD input format.

Why: Because this is the most efficient/best format to store separate AAC files, and it wasn't supported yet. In addition, reading AAC files is needed to be able to replace AAC audio tracks with custom audio when pushing RTMP out, a feature that is also part of this release.

Support for overriding AAC audio in RTMP push output

What: If a stream contains AAC audio, this feature lets you replace the audio in a stream with a looped version of an external AAC audio file. The original audio is not transmitted.

Why: This was built to support the use case that the stream may contain audio material that causes e.g. Youtube or Twitch to block the stream (e.g. copyrighted songs etc), and you want to replace the audio track with a placeholder for sending a "lite" version of the stream to these platforms.

Split config support

What: The ability to split up the config file by section, so that each section can be loaded from a different file on the system.

Why: This makes it easier to set up a generic config that can be re-used and synced across multiple installs, while keeping some of the config stored separately for each install that won't be overwritten. It's also possible to make part of the config read-only and other parts read-write by setting file permissions of their respective files accordingly.

RIST support

What: RIST is a semi-reliable transport, similar to Haivision SRT, Zixi and BRT. This means that it makes a connection that cannot be controlled (e.g. over the public internet instead of private networks) fairly predictable in behaviour. This adds support for RIST push output and RIST pull input, both in both caller and callee modes. Unfortunately our build system does not yet support building the RIST library itself, so this feature is currently only available if you compile MistServer yourself.

Why: RIST is rapidly gaining popularity, and there is still no clear "winner" in the reliable transport protocols segment today. Mist aspires to connect anything to anything, so we try to support as many relevant protocols as possible.

HEVC/H265 support in browsers

What: Most browsers don't support decoding of H265/HEVC video streams through any method (with or without hardware acceleration). A notable exception is recent versions of iOS, but for the rest mobile and desktop browsers alike generally won't support this codec. It is, however, possible to transpile libde265 (a software-only decoding library for H265/HEVC) to javascript and "force" software-based decoding of the codec in-browser anyway. This adds support for playing back H265/HEVC feeds through this method.

Why: H265/HEVC has a much better compression ratio than H264/AVC does, which can be critical in ultra low bandwidth situations. Unfortunately, software decoding of H265/HEVC is fairly slow - so this works best for low framerate or low resolution streams. The player does automatically skip frames to stay roughly at real time speed if/as needed, though - so even full resolution playback should be acceptable quality. There is another downside: the player currently only supports video, no audio.