There may be a lot to complain about with Java but it is insanely efficient when it comes to serving data out a network port. Wowza will scale well beyond most companies' needs, I've used their stuff for years and it is the one program that convinced me that Java actually can be very efficient for applications like these.
The setup seems rather similar to https://github.com/arut/nginx-rtmp-module which I've used in the past.
Perhaps they've bundled it with Nginx Plus without crediting the original authors? The project has a bit too permissive license for my taste.
The scalability of DASH/HLS comes from the fact that the video segments are just static files sitting on a http server so they can be cached and distributed with the many techniques for serving static files over http.
If you don't need the scalability of DASH/HLS you can use the WebRTC apis for low latency streaming, usually <1s. WebRTC can be used for client-server applications just as easily as peer-to-peer using a gateway like Janus[1].
Abusing WebRTC can be a viable option in latency-critical applications, but it brings a bunch of other concerns with it (browser support, no way to leverage a CDN, etc).
Streamers typically use something like [1] OBS, or [2] FMLE to stream their desktop or webcam to an RTMP ingest server like Twitch, Youtube, Streamup, Streamboat.tv, etc. Can even just use ffmpeg.
ffmpeg can do that fairly easily. To adjust the command in the example, you need to to take the srt file and pass it as a second -i input file. There's some helpful guides on how to do that if you search around.
The one thing I've always wanted from nginx-rtmp was the ability to pull from a source input instead of running a separate process with ffmpeg to push the stream over. If they could get that working my life would become less complicated quickly.
yeah I use nginx-rtmp for local video capturing (basically the functionality of a mirror) and the lag that needing to re-encode + chunk + starting off at the beginning of the chunks in the playlist introduces is meh. I think I'm going to transition my project over to WebRTC though, since the mobile device I'm using to display the video can handle that now.
Currently to get the stream from ffmpeg into nginx you have to do the following:
ffmpeg -re -i input_file -c copy -f flv rtmp://nginx_server_url:1935/app
It would be nice to have the ability to set up an app block that looked kind of like this
Using Representations[1] in MPEG-DASH is a more native way within the manifest to scale the quality / bandwidth requirements. There is some preliminary work and an issue open on one of the module forks which seems to get a little more tlc (imho) - https://github.com/sergey-dryabzhinsky/nginx-rtmp-module/iss...
video.js with the HLS plugin is fine if you don't need dash (there may be flash) but obviously that's very bare bones. It seems like there's things like ad support among other features you get with the suggested player.
I had the same wonder and through some digging found videojs-contrib-dash[1] which uses dash.js, referenced on the bitmovin site[2]. I'm going to give this a go and see what does and doesn't work.
57 comments
[ 3.3 ms ] story [ 119 ms ] threadhttps://www.youtube.com/watch?v=1boJWioxsWc
If you don't need the scalability of DASH/HLS you can use the WebRTC apis for low latency streaming, usually <1s. WebRTC can be used for client-server applications just as easily as peer-to-peer using a gateway like Janus[1].
[1] https://janus.conf.meetecho.com/
http://www.gpac-licensing.com/2014/07/09/lowering-dash-live-...
http://www.coderag.com/hooli-lays-off-hundreds-of-employees-...
How do I create an RTMP stream in the first place?
What camera can/should I use? Can I use webcam? Is there an iOS app that can do it?
What hardware/software is needed to create this RTMP stream that I'll be pushing to nginx?
[1] https://obsproject.com/
[2] http://www.adobe.com/products/flash-media-encoder.html
For example, I want to mux a subtitle stream, with the real time clock in it, into a video stream.
e.g. https://en.wikibooks.org/wiki/FFMPEG_An_Intermediate_Guide/s...
Could you elaborate on this? I don't understand what you mean by 'push' and 'pull' in this context.
It would be nice to have the ability to set up an app block that looked kind of like this
application source1 { exec fffmpeg -f decklink -i 'DeckLink Quad (1)@8' -f flv rtmp://localhost/app/$name }
ffmpeg is still pushing the stream to nginx, but nginx is in charge of starting that process.
https://github.com/arut/nginx-rtmp-module/wiki/Directives#ex...
Most streaming servers are bloated and slow. If nginx can do for video streaming what they've done for web servers it would be great.
Uhuh. Suuuure. :p
[1] https://www.brendanlong.com/the-structure-of-an-mpeg-dash-mp...
[1] https://github.com/videojs/videojs-contrib-dash
[2] https://bitmovin.com/mpeg-dash-open-source-player-tools/