Question / Help RTMP stream without buffer ?

[thasmog]

Hi im using legacy OBS and video source plugin.
My issue is that the plugin adds 3000ms buffer to my local RTMP stream. I have to use different program to get the stream in with 0 buffer.

Im using Action! With NVENC (Id like to use quicksync but action wont work with my gf pc in quicksync setting) in first pc to stream into second. I capture stream with vMix and add it into OBS. I'd like to skip the vMix part becose its using 20% of my cpu. With this setup i get 0.5sec delay.

Also i use MonaServer.

Tomorrow i test to capture second mix with 3.5 splitter and line in. Is there better ways ?

 

[Harold]

Tried with obs studio and media source instead of classic and video source?

 

[FerretBomb]

You can experiment with turning down the scene buffering time and buffer settings, but it can result in significant issues if your setup and network aren't tuned for rock-solid throughput and performance. Using a buffer of 0kbps with VBR could work for slinging across the network, but will be peaky and could cause playback issues; a low but still non-zero value would be better. I would recommend not going below 35ms scene.

This is not advised for anyone who is not doing a LOCAL RTMP STREAM. Meaning if you are streaming to Twitch, don't do the above unless you are very technically experienced with the RTMP protocol, as you WILL mess up your stream.

 

[thasmog]

Tried with obs studio and media source instead of classic and video source?

Yes i did. Dont rembeber the delay there but i prefer legacy becose studio wont support 3gen QuickSync.

https://youtu.be/PdXbIHt12IQ <- here you see output. Setup is not ready and wewerent even home. Tomorrow i test again at home.

 

[thasmog]

You can experiment with turning down the scene buffering time and buffer settings, but it can result in significant issues if your setup and network aren't tuned for rock-solid throughput and performance. Using a buffer of 0kbps with VBR could work for slinging across the network, but will be peaky and could cause playback issues; a low but still non-zero value would be better. I would recommend not going below 35ms scene.

This is not advised for anyone who is not doing a LOCAL RTMP STREAM. Meaning if you are streaming to Twitch, don't do the above unless you are very technically experienced with the RTMP protocol, as you WILL mess up your stream.

I had not any proböems with todays stream :0 i got pretty good modem with 1gb link and with dual wireless dos pretection and everything else. It even can run as a file server or media server. Got it free and its pretty awesome. Im sure it wiöl handle it.

 

[FerretBomb]

The disclaimer at the end was more aimed toward newbies who might skim the advice and apply it to their direct-to-Twitch stream, which will end poorly. :)

 

[thasmog]

Here is source for the plugin im using. It would be awesome if somone who knows this thing can add buffer feature:https://github.com/kc5nra/VideoSourcePlugin

Also i tryed to find something in the code but nothing. I think its codecs default to set it 3000ms

 

[thasmog]

Up ?

 

[Jack0r]

The original creator of the plugin is quite busy in his real life as far as I know. And the main development has moved over to OBS-Studio.

 

[thasmog]

The original creator of the plugin is quite busy in his real life as far as I know. And the main development has moved over to OBS-Studio.

You cant set bufger timd in studio either. If im not wrong. Also OBS Studio wont support 3gen quickync. (It uses some microsoft things)

 

[RADRaze2KX]

Thasmog, did you ever find a solution to this problem? I'm trying to reduce the latency down an RTMP server as well and am having the same issue you did. I've experimented with various methods of RTMP streaming thinking there was a problem with my server, when in fact it's all the clients I've been using. I had never heard of vMix until I saw this thread and it works great, but not everyone I will present this solution to will be able to use it with such a heavy CPU usage. Here are the results of my testing:

Running a simple stopwatch and then taking a snapshot when all the watches were going resulted in the differences in their delays. Subtracting the lowest time from all numbers accounted for my reaction speed, and then listing them in order of smallest to largest number gave us the fastest to slowest method. vMix18 had a 1-second delay from the live stream test, so adding 1 second to all the times gave me their actual delay times. Have you found a different way to capture the RTMP feed into OBS Classic?