Ultragrid via NDI6

[finnjaeger]

So when it gets a 422 signal it would tautomatically zero-conf do 422 h264? i assumed i needed a bunch of extra ffmpeg lines in my config…

ill check logs tomorrow.

[cnoellert]

@finnjaeger

…see this is perfect.Any specific dope ultragrid argument lines to make h264 work like a boss for something solid-enough like 4:2:2 ?For me h264 never was really that awesome without going the full beans SRT route (i had this working before a long time ago)and yes you are of course completely right how bad jpeg2000 is! changed my first post to prores as that does work great for me.

One thing that’s worth considering is actually picking up the signal on another machine dedicated to streaming the UG signal outbound. Depending on what you’re looking to do with regards to codecs certain platforms can be better suited to this task than the platform you’re running Flame on.

For example, for all of my on-prem sessions I stream prores to the broadcast and edit suite “client” displays (There’s 3 client displays, don’t know why…) by encoding on a Macmini client which is using Apple’s Prores-VideoToolbox encoder which reduces latency to idiotically low-levels while not eating up huge amounts of system resources. I used to do it via HDSDI but have no moved over to NDI. The signal is kinda fat but that’s not an issue and you can beam that shit everywhere on the local network like a crazy person. Likewise for H265, but nvenc hardware encoding on a Nvidia GPU also has stupidly low-latency (if you’re fine with YUV which at 15mbit I am).

In those scenarios you spin-up UG on a machine on the same network has your flame host and use it’s resources instead. To that point, you can have one machine that’s picking up the signal for NDI and beaming it to your broadcast monitors over UG, and then using LBS to pickup the same signal and push it to Louper. Once it’s NDI that shit can be everywhere.

[finnjaeger]

yea Ive got a dedicated macMini with a m2 to do the encoding stuff also on the client I have a dedicated machine just to keep it neat!

[finnjaeger]

i do LIKe my NDI stuff and thats all great , absolutelty-

Having a way to just by installing a single tool enables you to get a nice signal down to your home machine is worth its weight in gold, you are freelancing at a place and just need to get smooth playback in high quality? they might only run teamviewer or anydesk even…

this is a godsend, takes 5 min to make it work or as alan years of develeopment to find the perfect thing for his clients.

[ALan]

@finnjaeger

So when it gets a 422 signal it would tautomatically zero-conf do 422 h264?

That is the idea, and how it works with SDI, although you at least have to specify the codec of course. I do everything in headless Ubuntu server environment, but the GUI should be even easier.

[finnjaeger]

ill give it a go and check the logs then i just assumes it would just pick a random h264 default preset

[cnoellert]

On an M2, provided you’re running Sonoma and UG 1.9, the setup is fairly straightforward. For encoder use libx264, set the bitrate to whatever and the subsampling to 422. Argument-wise that translates to:

-t ndi:url=192.XXX.XX.XXX:5961 -c libavcodec:encoder=libx264:bitrate=15mb:subsampling=422 -s embedded --audio-codec u-law destination-ug-instance 

This will do v210 door to door which is supported client-side by BMD.

[ALan]

Why u-law? Isn’t that a low fidelity codec?

[ALan]

@finnjaeger

random h264 default

Not random, UG generally understands the input format and chooses the relavent colorspace and stuff. But I’ve only used SDI inputs.

[ALan]

@cnoellert

…you’re fine with YUV which at 15mbit I am).In those scenarios you spin-up UG on a machine on the same network has your flame host and use it’s resources instead. To that point, you can have one machine that’s picking up the signal for NDI and beaming it to your broadcast monitors over UG, and then using LBS to pickup the same signal and push it to Louper. Once it’s NDI that shit can be everywhere.

If you are onsite, and outputting NDI from flame, why not just puts an Apple TV at each client monitor and use the Sienna app to grab NFI directly from flame. Skip the whole UG complexity for internal purposes?

[cnoellert]

Sure, but for me on-site means on-premise at a client location or the space where we do sessions. Not where the kit is located. The Sienna stuff works great though.

[ALan]

Ah… understand.

[ALan]

I was thinking, anyone interested in doing a Zoom dedicated to this topic? I’d be up for a show’n’tell.

[cnoellert]

@ALan

Why u-law? Isn’t that a low fidelity codec?

Just a miss when testing the line in gui…

It’ll work but yeah not amazing.

[cnoellert]

@andymilkis ears begin burning…

[dave]

This is super interesting to me. I would love to learn more about this.

[ALan]

I just enabled NDI in Flame 2025 to throw it thru ScopeBox, and it still can’t do RGB 10/12bit 444, and only can do YUV 709 422. I’ll be spending zero seconds more contemplating NDI.

[finnjaeger]

yea…

well you aint going to get a sdi out from a aws/cloud machine

[Taliessyn]

So one of our use cases at Replayboys is the fact that I drive a Flame in Hamburg while sitting one country over in Amsterdam. In our short time testing ultragrid its been pretty great.

I do tease Finn quite often about the fact that Germany’s internet is strangely bad while here next door in this tiny ass country you can get 4gbit for next to nothing cost wise.

[ALan]

Unless it is private cloud.

@finnjaeger

well you aint going to get a sdi out from a aws/cloud machine