NDI

I'm just now looking into NDI

And I installed the NDI tools. Looks pretty nice.
https://www.newtek.com/ndi/tools/#download-tools

I noticed that the tools get distributed with NDI plugins for adobe AE and Premier.

I wonder what strategy BMD has regarding NDI.

Their strategy is supporting OFX.
Much of that is competition anyway.

Their strategy is supporting OFX.

no -- resolve uses in fact a modified version of OpenFX, which violates standard compliance and compatibility!

and although the actual extensions would make a lot of sense for other applications as well, it's more or less characteristic for BMDs general strategy, do not use their voice and influence in the relevant standardization boards to extend the OpenFX standard in a useful manner, but simply establish proprietary solutions.

but NDI does not have much in common with OpenFX. it's not a videoFX plugin API, but a video transport/communication standard. and it's indeed a very useful and innovative approach, which saw a lot of support and adoption in the industries since it's presentation two years ago.

but this kind of exciting innovation and open accessible SDKs represent the strict opposite to BMDs general principles, which are much more focused on forcing all users to purchase legacy broadcast devices and rigidly closed software solutions, which aren't accessibly/extendable by third party developers...

NDI is currently working well on Windows but Mac support for anything other than Adobe products is almost non-existant. Newtek provide codecs for their NDI recording software but they are already out of date on Mac. I have been in discussions with them about updating their codecs but they don't seem to understand the issues. In a Windows environment it's quite a good toolset, not for Mac or Linux. I would expect that BMD would wait until all platforms can be supported properly.
BTW I have been using NDI to distribute scoring graphics in livestream production for sporting events and we can't do any post-event graphic amendments in a Mac edit suite due to these issues. It is a really great technology, but still early days in terms of development.

no! -- i wouldn't see the support of NDI on non-windows systems as utterly insufficient.

it's already accessible by ffmpeg (>3.4) and all application which use its libraries for media i/o on all platforms. and it's also supported by media frameworks like MLT and Cinder and popular end user applications like OBS.

IMHO that's a quite impressive support for such a relative new solution, but it really hit the nail and represents a very promising approach/improvement to bridge the gap between ordinary computer infrastructure and the needs of more flexible and modern ways of video transport.

but i agree with you, that the multi-platform support by NewTek and its tool box of NDI standard applications could be indeed much better. and also the terms of access to the NDI SDK are unfortunately not optimal resp. not free enough for a more satisfying support and redistribution in many linux distributions, which try to take the question of 'free' software and licensing issues rather serious.

We may need to respect BM‘s business model here: they give us pretty good software for free and very powerful software for a really low price. I don’t think they can feed the developers and their families from these revenues alone. So, they restrict that software to their own hardware as much as they can. Maybe something similar to NDI is in the works?

Surely Newtek can simply release an NDI OpenFX plugin? I don't see how BMD needs to be involved here.

Surely Newtek can simply release an NDI OpenFX plugin? I don't see how BMD needs to be involved here.

i think, you do not even have a vague idea at all for what kind of practical purposes NDI and OpenFX were invented and why it doesn't make any sense to intermix this utterly unrelated approaches.

NDI is a connectivity solution, which allows you to utilize an ordinary local computer network for video streams. it's therefore more a kind of alternative to SDI or HDMI cabling, but especially focused on live video production and distributed monitoring, because the signal is (loosy) compressed in a similar manner as in Prores/DNxHR/HQX files. otherwise -- i.e. for uncompressed video transport over IP networks -- you would need very expensive advanced nd uncommon network infrastructure. NDI is therefore just a compromise to utilize much more affordable existing computer hardware in an acceptable manner resp. image quality and negligible latencies for practical video transport in more common envirionments.

OpenFX on the other hand, was invented as standard for video FX plugins -- or more precise: film and video related image processing, because it does not support sound/timecode/metadata information. in contrast to NDI it isn't optimized for any form of realtime constraints, but more oriented towards the demands of perfect image processing. sure, in some cases the OpenFX hosts may also utilize various technique to archive satisfying realtime processing (e.g. frame droping, prefetching and caching, distributing the load etc.), but that's not one of the main goals or common requirements of typical OpenFX usage.

in fact it would be even possible to write an OpenFX plugin, which could utilize at least some aspects of typical NDI capabilities. but that's more a theoretical point of view, because most OpenFX host -- and resolve in particular -- do not support the OpenFX generator context in a sufficient manner. otherwise we could use this kind of interface also for file support extensions (e.g. wrapping ffmpeg) -- although without any sound/timecode/metadata support, which again has to be seen as unacceptable in practice.

if resolve would support NDI in a more adequate way, it would have to be done more like virtualized decklink devices -- a kind of of alternative input and output possibility over the network. but i wouldn't see it as much as a competing solution, it's more a complement to more traditional forms of video connectivity. in most real world scenarios it's simply not useful to utilize this kind of computer mediated video transport, but in those rare cases, where you really need this kind of remote access over the LAN, it's indeed a very nice solution to solve challenges in an extraordinary elegant manner.

On the Mac side of things

Syphon and it's offshoots and variations
do support BMD cards and boxen

http://syphon.v002.info/
big list of supported apps there

for sure check out
"Black Syphon.app"

Syphon has been a solid performer for many years
used in the arts / events / live venues for years
very rare to see in Post Production

wh

syphone and other simple framerserver solutions but also some webcast solutions are in a few regards rather close to the capabilities of NDI, but i wouldn't see them as an equal alternative. NDI is still much closer to a low budget "AV over IP" alternative to SMPTE 2022-6/2119 and similar professional standards. and it's not only a pure software solution, but also supported by many hardware devices (e.g. cameras, external converters, live video switches and streaming boxes.)

see: https://en.wikipedia.org/wiki/Network_Device_Interface

i think, you do not even have a vague idea at all for what kind of practical purposes NDI and OpenFX were invented and why it doesn't make any sense to intermix this utterly unrelated approaches.

I know perfectly well what NDI is, I just wasn't aware that OFX was as limited in functionality as you describe.

I wonder how an OFX plugin effect could provide an NDI video input or output for Resolve. I would not have thought it possible.

I wonder how an OFX plugin effect could provide an NDI video input or output for Resolve. I would not have thought it possible.

in some other video applications (e.g. Natron) all I/O is realized by OpenFX plugins (e.g. by openfx-io). this approach does work very well for compositing, FX or pure color grading applications, where you do not need sound, but it isn't an optimal solution for NLEs and general purpose video processing solutions. and in case of NDI it would be also a very questionable choice, because this kind of access is usually only used for different kinds of unseekable live video sources resp. AV streams shared in realtime between different devices on the local network.

This has been a very interesting thread. Thankyou Martin for sharing your considerable knowledge on the topic.

I haven't looked at the Adobe NDI plugins yet, but because Premiere and Resolve are both NLE programs, I thought that however Premiere has used NDI, that Resolve could too.

I looked at some of NewTek's marketing material and it looks like they've implemented NDI in NLEs simply for the purposes of realtime preview across the network so that work can be reviewed in realtime over the LAN (and perhaps internet if link is fast enough). So basically NDI would be in the context of an enterprise environemnt/workflow.

So when you are in resolve and you playback your program preview using CTRL-F, you get fullscreen. Maybe as well, you could cofigure it to be sent to a broadcast address so that other team-members could preview the edit in realtime whilst chatting over the phone or voip.

I watched a

with one of the NDI lead developers and he said that Linux support is in the works.

The recent Windows 10 update rendered my laptop useless for video editing because of the massive performance hit. Even with a reformat/re-install, it is still slow. I installed UBUNTU 18.04 LTS and basically got the usage of my laptop back as a workhorse. I since installed Resolve free on it using the makeresolvedeb script.

http://www.danieltufvesson.com/makeresolvedeb

What a wonderful human being Daniel Tufvesson is for making this public open source. There's no way I would be able to figure this out on my own and I am only comfortable using Ubuntu LTS for pro work.

I don't think OBS has native NDI support yet like vMix does. You can use it in OBS via the virtual webcam driver that is part of the NDI toolkit. Not sure how much latency that method introduces but I think it's significant.

Most linux distros have a way to incorporate non-free stuff via a separate repository so I don't think that should be an obstacle to the Linux implementation of NDI.

I paid the AU$31 and installed the NDI app on my iPhone 6 and its pretty amazing functionality to have a roving camera over wifi for a live feed.

I think that the democratization of live television by way of social streaming media is a paradigm shift. Like with the invention of the Internet, HTTP came along and suddenly it exploded. I think NDI is analogous to the invention of HTTP. Sure, there were other things doing basically the same thing before, but NDI does it in a way that is reliable and very focused on the new market segment of "garage streamers".

I guess BM's product set that could use it the most are they ATEM switches. There's another Australian company that is innovating on hardware NDI and perhaps they should team up. Check this out (highly recommended).

http://www.bird-dog.tv/

This product has the potential to demote the ATEM switch to an upstream device so that NDI streams can be incorporated into the stream.

I don't think people appreciate that video streams are more than just things that camera produce, a desktop could be a video stream, a real-time telemetry graphing app could be a live video stream. All of these are readily publishable as NDI today! I don't think the ATEM switch can handle that being done in the software domain (cheaply and replicatably).

We are definitely in time of re-inventing video and audio transport and quite obvious choice is treating video as "normal" data and use existing network.
There are companies which already send 4K (even uncompressed) signal from London to Europe over private dark fibre links (with <1 frame delay) and technology around it isn't that crazy sophisticated (although it can be if you add encryption and advanced error correction).

If anyone is interested, there is a free course on udemy. It's not great, because the guy repeats himself a lot. Jump to section 3 and go from there.

https://www.udemy.com/newtek-ndi/learn/ ... 66?start=0

We are definitely in time of re-inventing video and audio transport and quite obvious choice is treating video as "normal" data and use existing network.
There are companies which already send 4K (even uncompressed) signal from London to Europe over private dark fibre links (with <1 frame delay) and technology around it isn't that crazy sophisticated (although it can be if you add encryption and advanced error correction).

I was watching this video

And shortly into it, he shows the guy a mini theatarete where staff have a 4K tv installed where they can watch someone on the other side of the world grade footage in realtime.

Many bigger studios use remote grading (not talking about "BM way" with duplicated data, but realtime feed). It can be done at different quality/complication level (key point is low delay). Typically this is going through hardware boxes with jpeg2000 encoders, but this is quite expensive way.
NDI was rather designed for local networks (as far as I understand), but with todays fibre links nothing stops you to use it over VPN. Not sure about its protocol (not sure if it has good error correction), but there are also open source projects with very advanced error correction (+ encryption if needed), so this can be done also cheaply if you really need it. At the end it's mainly a matter of how good link you have between A and B. With good link you can push 20Gbit/sec+ if needed, so even 4K uncompressed.

Many bigger studios use remote grading (not talking about "BM way" with duplicated data, but realtime feed). It can be done at different quality/complication level (key point is low delay). Typically this is going through hardware boxes with jpeg2000 encoders, but this is quite expensive way.

yes -- there are really impressing solutions available for this kind of remote access and screen mirroring available. if you work on the linux platform, Xpra is such a [free] software. it can even utilize the hardware acceleration capabilities of nvidia graphic cards for more efficient compression and the delays during transmission are impressive small.

NDI was rather designed for local networks (as far as I understand), ...

it was indeed mainly designed for use in LANs, with features like automatic device discovery via Bonjour zero-configuration networking etc.

but with todays fibre links nothing stops you to use it over VPN. Not sure about its protocol (not sure if it has good error correction), but there are also open source projects with very advanced error correction (+ encryption if needed), so this can be done also cheaply if you really need it. At the end it's mainly a matter of how good link you have between A and B. With good link you can push 20Gbit/sec+ if needed, so even 4K uncompressed.

in principle that's possible, and this kind of long distance links were also successful demonstrated in practice by NDI customers. but it's not very well optimized for this kind of purpose. in this case it's usually recommended, to use a gateway, like sienas NDI cloud, for more efficient transport.

but in general it's important, to see, why NDI isn't only just another arbitrary remote access or screen mirroring solution. it's really a protocol, which supports those few very important features of professional video transport (time code, high bitdepth and sufficent color metadata information, low resp. compensated delays, audio synchronization provisions) in a satisfying manner, which isn't the case in many other networking and video delivery solutions, although they may have other advantages again...

Plus they got alpha channel support.

Here is a simplistic demo of the Adobe implementation. I assume they have very fast hardware to support this with alpha channel and fast laptops which are optimised (with not much else running in the bg).

They also use an example of sending live graphics (lower thirds) to a live broadcast, but I have my doubts as to the reliability of this even though I am not an expert. I think such a use-case scenario is ambitious.

I haven't looked at their SDK, but I assume there is enough there for BMD to implement an NDI option for program preview.

Being that everything about NDI implementation hinges on network infrastructure, this presentation is very helpful.

NDI is currently working well on Windows but Mac support for anything other than Adobe products is almost non-existant.

This simply isn't true.

There are lots of NDI products for Mac, and a growing number for Linux
https://en.wikipedia.org/wiki/Network_Device_Interface

NDI is currently working well on Windows but Mac support for anything other than Adobe products is almost non-existant.

This simply isn't true.

There are lots of NDI products for Mac, and a growing number for Linux
https://en.wikipedia.org/wiki/Network_Device_Interface

The main issue is video codec support for post-production. The Newtek codec is still mpg2 based and the alpha channel does not work on OSX after the Apple decision to drop the mpeg2 component.

It would be good for Resolve to have an NDI in and out though!

Are you sure?
It's just another intermediate codec, more like GV HQ/HQX. It's not decoded by mpeg2 decoder in the system, but you need to install it:
https://www.newtek.com/downloads/

Other than this: you don't need support for the codec itself as normally it's used for transmission only and your sending/receiving app decodes internally. It's only in some special cases when you would need to work with files saved in their codec.
You can try it. There is fee Adobe plugin. Then on other side you can get trial off an app (http://www.sienna-tv.com/ndi/ndi-outlet.html ) which connects to your stream and sends it to BM card. All quite simple.

Resolve plugin should be quite easy to develop, specially if Nobe Display already exists.

It's just another intermediate codec, more like GV HQ/HQX. It's not decoded by mpeg2 decoder in the system,...

the utilized codec is in fact very similar to other simple intermediate formats like DNxHD and Prores, too.
https://wiki.multimedia.cx/index.php/SpeedHQ

It's fairly simple and fast. Only issue is that most NDI based application are (I think) 8bit, even if Newtek says 10bit or even 16bit is possible:
https://support.newtek.com/hc/en-us/art ... g-Decoding

It's about 100Mbit for HD, so not crazy good quality, but decent enough as network based stream.

Now, in 2021 we have the OpenFX NDI output plugin (Nobe Display with NDI support - https://timeinpixels.com/nobe-display/).

Many apps have NDI support now and list keeps growing.
Nobe way is a bit of "hack" (not that big though). It would be nice to have native NDI input/output in Resolve.

Avid MC has native support for NDI. You can broadcast your video playback over apps like Zoom. Quite useful.

It's fairly simple and fast. Only issue is that most NDI based application are (I think) 8bit, even if Newtek says 10bit or even 16bit is possible:
https://support.newtek.com/hc/en-us/art ... g-Decoding

It's about 100Mbit for HD, so not crazy good quality, but decent enough as network based stream.

From my tests, if you request the highest quality stream in NDI API - 16bit signal format will be used on Windows. On macOS it will always default to 8bit. Haven't found a way to get anything more than that on mac.

10bit is growing. There are new hardware NDI boxes which support 10bit. Flame support is also done at 10bit if I'm correct.

Avid MC has native support for NDI. You can broadcast your video playback over apps like Zoom. Quite useful.

Actually, not really. Broadcast requires a set of NDI Tools and a webcam emulation. Because Zoom, for example, does not support signal sources in the NDI format.

Avid MC has native support for NDI. You can broadcast your video playback over apps like Zoom. Quite useful.

Actually, not really. Broadcast requires a set of NDI Tools and a webcam emulation. Because Zoom, for example, does not support signal sources in the NDI format.

?? NDI is an input option for Zoom.

https://support.newtek.com/hc/en-us/art ... m-on-MacOS.

It is also supported on Windows.

We have a show doing this now.