**Testing NVIDIA's Broadcast Technology with Modern Warfare 2019**

I recently had the opportunity to test NVIDIA's broadcast technology, which is designed to enhance streaming performance and provide better image quality for content creators. I started by testing it out on Modern Warfare 2019, a game that I've been playing a lot lately. As expected, having both voice and camera enabled during streaming would definitely have an impact on how smooth and responsive the game felt compared to normal streaming.

When testing the broadcast technology with Modern Warfare 2019 at 4K resolution, I found that it had a significant impact on game performance. The broadcast effects had a 14% impact on game performance, while streaming normally with reasonable settings on the NVIDIA GeForce RTX 3080 resulted in a 23% impact. However, there is still an impact on both cards, and this will be a trend throughout all of the games that I test.

On the other hand, at 1080p 240, I found that using the broadcast technology had basically no impact, while enabling max quality in ending settings had a greater impact on the GeForce RTX 3080. The NVIDIA GeForce RTX 2080 suffered a 20% performance impact when using the broadcast software.

**Comparison of Broadcast Technology Impact with Different Games and Resolutions**

When testing the broadcast technology with Apex Legends at 4K resolution, I found that the GeForce RTX 3080 had a 16% performance hit moving from normal streaming to streaming with broadcast. In contrast, the NVIDIA GeForce RTX 2080 had a 13% performance impact without the camera on, but when both camera and voice were enabled, it went up to a 30% performance hit.

At 1080p 240, I found that the GeForce RTX 3080 only suffered a 10% impact from the broadcast effects, while the NVIDIA GeForce RTX 2080 had a 19% impact from just the camera effect and a 30% impact when both were enabled. These results show that even with the same resolution and frame rate, there can be significant differences in performance depending on whether or not the broadcast technology is used.

**Testing Broadcast Technology with Rainbow Six Siege**

When testing the broadcast technology with Rainbow Six Siege at 4K resolution, I found that the GeForce RTX 3080 saw a 15% impact with both effects. In contrast, the NVIDIA GeForce RTX 2080 only had a 12.5% impact from both.

However, when comparing these results to the same test at lower resolutions and frame rates, there were significant differences in performance depending on whether or not the broadcast technology was used. At 1080p 240, I found that both cards had only a 6% impact over normal streaming, so they tied in this test.

**Conclusion**

Overall, NVIDIA's broadcast technology is phenomenal and has the potential to greatly enhance streaming performance and image quality for content creators. However, there are still some limitations to consider, such as the fact that it can have a significant impact on game performance at high resolutions and frame rates.

One of the biggest challenges with this technology is that it relies heavily on real-time processing power, which means that it can be a bottleneck in certain situations. This is why I think that developing a video editor plug-in for NVIDIA's broadcast technology could be incredibly valuable. Imagine being able to apply these effects directly within a video editing software, where you have all the time in the world to process and optimize your results.

As for future development of this technology, I'm excited to see what's in store. With more training data and testing, NVIDIA can continue to refine and improve their broadcast technology to provide even better performance and image quality for content creators.

**Recommendations**

If you're interested in trying out NVIDIA's broadcast technology for yourself, I recommend checking it out. It's free to try, so there's no risk involved! Just submit your own sample footage and let them know what you think about the results. If you have any feedback or suggestions, be sure to share them with the community.

As always, if you enjoyed this article, please hit the like button, subscribe on Nebula or YouTube for more tech education and stream content, and I'll see you next time!

"WEBVTTKind: captionsLanguage: enall right today we're talking about the nvidia 30 series graphics cards again except not from a hardware perspective that video was yesterday or earlier today depending on when you watch this or when i post everything and how things are going today we're talking about the nvidia broadcast software this is software that they released alongside these graphics cards to filter out your voice for background sound or your chat member's voice for background sound as well as remove your background put you in front of any background or blur it out or whatever it's really cool software and i could not fit everything i wanted to cover in that one video so i'm making a second video about it let's go so this is nvidia broadcast this was formerly rtx broadcast but they're renaming it even though this specific software does actually require you to have an rtx card unlike the rtx voice that they officially open to all gtx cards as well this was shown at twitchcon 2019 and seemingly never heard from again until now it combines rtx voice with a virtual green screen app both built on lots of ai training models and provides some really powerful tools for streamers it still needs some work but it's a fantastic start and i can't wait to see where it goes and also i just can't wait to see when it gets added to video editors for post work because i think there's some huge potential there just to be clear this is not analyzing and sending back your video or audio to ai to then judge on the fly and change and learn as you go it's trained in advance and improves as more training can be done it was a weird myth that was getting spread when the original rtx voice launched i don't know what happened there this is of course a work in progress from a small team please give any feedback or contribute samples for training at broadcast.nvidia.com feedback all right let's walk through the software nvidia broadcast is this little app with three tabs two for voice and one for video let's start on voice formerly rtx voice though officially opened up to all nvidia cards with varying results this past week the audio portion of nvidia broadcast is designed to help filter out unwanted noise from your microphone or from voice chat if you have the hardware for that it has a quirk this is built on ai training of voices and microphones versus environmental sound to filter out just the voice in a way that at least in my previous round of testing i found to be way higher quality and more natural sounding than the results from most real-time noise suppression or noise removal options it's quite impressive as you can see here here's a vacuum the vacuum is running the vacuum is running the vacuum is running the vacuum is running the vacuum is running the vacuum is running the vacuum is running you'll always have some impact on the vocal sound due to frequencies matching between whatever background sound and your voice but cutting out a vacuum is crazy and yes i was the first one to do this on a youtube video everyone else seemed to just rip me off and then got all the credit oh well this does work best with cardio and microphones such as desktop streaming mics or headset mics it works less well on boom mics and even worse on omni-directional mics such as lav mics i will have more testing of this in the future i just haven't had time yet for microphone settings choose your microphone device as an audio source from the list toggle on the noise removal and set the strength of the filter they even provide a neat little tool to record and play back a clip of your voice to see how the filter affects it so you apply it normally listen to it then up you know increase the strength listen to it again until you get to a sweet spot that gets you what you want pretty straightforward that's it for speakers you do the same thing but with an audio output device this is for voice chats or conference calls if someone has a lot of background noise or keyboard sound or something really obnoxious they have simple background sounds available to playback as wave files so you can hear how the strength of the filter affects them to get you know a sweet spot of where you want so that is not too overbearing however a problem you may run into with the speakers is that unless you're purely doing a voice chat setup for your audio you don't have any other sound going on you'll need a dedicated output for your voice chat to run through because otherwise it'll filter out every other sound coming through your system your game sound your music your video you're watching anything so if you're using a physical hardware mixer or separate without separate output runs or something like voice meter software that's fine but if you have everything running to one device that will present a problem it's also a problem for the go xlr as the filter sound is actually presented as a separate output device that you then listen to not added back into the original output device that wouldn't work so then you can't add it to your go xlr the elgato wave microphones however do not have this problem because you can add devices to the software then you have the camera beta tab here you select your camera source the resolution and frame rate to run it at your effect and a preview of it happening currently it mostly supports uvc devices only so devices that present as webcams this is mostly limited to actual webcams but of course it also includes capture cards that present as uvc devices as well including the cam link the magewell usb cards the extreme cap uvc the hd60s plus as with the cloner alliance flint 4kp plus i reviewed as well as the 15 cant link that i made a video on and has since spread over the summer like crazy that's also uvc and so with those uvc capable capture cards you can use your real camera with hdmi like i am here in order to run it through nvidia broadcast so that's fine more expanded capture card support will be added in time and i really really really hope and i'm gonna hammer home that i really want them to add support proper support for black magic capture cards as these present differently than most capture cards and are usually left out but literally all of my camera outputs run through black magic cards and i really don't want to get left behind because they don't want to support it the resolution and frame rate that you choose impacts the workload of the neural network related cores on the graphics card in order to produce the image so if you're only doing a small face cam view in the corner of your stream you'll want to run it at a lower resolution so that it will have to work less and impact your game performance even less or have better performance for the processing overall i have not found lower resolutions to significantly impact the quality of the effect at all uh but there will be some differences by nature of there being less detail you know being resolved it does run at 4k though so that's pretty cool there's four effects to choose from for your camera back out background blur background replacement background removal and auto frame unfortunately auto frame cannot be run at the same time as other effects background blur is fairly self-explanatory it blurs out everything but you and gives you that faux bokeh depth of field look that i have right here this isn't faux though this is the real thing if i get close you can see it change that's what a wide aperture lens does it's especially convincing to use to enhance an already somewhat blurred out background from a lens from a mirrorless or dslr camera so if you have a real camera and it has some degree of background blur but not as much as you would like you can enhance that with this feature and it looks a lot more natural because any kind of imperfections are going to blend in a little bit more because it's already somewhat blurred out so that's pretty cool you all know about now that my office is also a living room where most of our baby furniture is since that's where he spends most of his time and a garage with my server rack and workbench behind it so whenever i need to record videos in that space background blur is actually really useful here there's a slider to control the intensity of the blur background replacement removes your background entirely and places an image behind you i use jensen's kitchen provided in the nvidia community pack linked below i recommend making sure that your image is decently blurred you can add that blur effect in photoshop or or critter or whatever as you like um as a super sharp image especially one that's more sharp than say your webcam is very distracting and much less appealing to look at background removal just removes your background and gives you a transparency compatible video that you can add to your streaming program and add your stream without using actual chroma key so why would you use background replacement over removal well for video calling and direct recording of video or vlogging scenarios where transparency won't be maintained the full removal gives you more control to manage in your streaming app but in a video call or direct recording all of that transparency is lost so you need to already have something presented behind you so that's where that comes in handy auto framing is pretty cool it will zoom in a bit and otherwise move your video feed around based on panning the shot around to keep yourself focusing in the center of the frame regardless of how much you move or shake or whatever you get a slider to control how much the initial zoom or crop factor is so that it has more room to work with within your frame so i recommend using a higher resolution image with that if possible um it's pretty cool some people like it unfortunately you can't do this and remove your background at the same time i realize that could look pretty silly but some people might want to you can't nvidia broadcast lives in your system tray and gives you a quick and easy right-click menu option to toggle camera speaker and microphone settings on or off in case they're impacting your game too much which i kept having having happen in modern warfare or if you have any issues or whatever you can quickly toggle them off next let's talk quality we're going to focus on the video side of things as i've already talked about voice in multiple videos at this point i'll have my samples linked in the video description overall the results from nvidia broadcasts for the background adjustment are pretty great depending on what's in your background of your actual shot the results can be beyond impressive when it's really working it feels revolutionary some of the clips where i'm only really lit by my computer monitors and using a crappy c930e webcam are so incredibly impressive it's been hard to keep keep this a secret until the embargo lifts but it's not perfect it does struggle with some of the same artifacting issues that most virtual background removal software do struggle with i have found darker backgrounds that work better than lighter ones as too many bright colors or shapes behind you can cause chunks of things to be detected along with you also having a second person behind you can confuse it a whole heck of a lot it's built for one streamer for example in terms of how it works i've been using it on my gaming side where i have much less going on in the background it's just our desk and the wall in the kitchen it's much further away that gets a lot better results especially with the lights off than my full living room background that i usually have with all the lights on as there's so many big shapes and colors and it just kind of confuses the the detection what's impressive is that when it is detecting right the edges of the cut around you look completely natural you can't even tell that there's there was ever something back behind you in the first place and you get to avoid the fuzzy green outline and color cast that a green or blue screen would cause however chairs are a problem the team had only a limited number of chairs behind users to train ai on and my primary chair the ergo chair 2 in gray confuses it even more there's no chair detection toggle like some options so you you know if the chair is always left in but if your chair does not look like what it expects you could have some issues it seems to prefer black chairs so swapping to my black respawn chair for the most part works better or if you have a mid-back chair like the one i'm sitting in right now it works even better because for the most part you're going to be blocking it with your body if your camera is angled down too far and it shows like your arm rests and stuff like that that could cause a lot of weird artifacting too because it has really like you can't expect it to train for that so you'll probably want to cut crop off the bottom of that feed a little bit so i tested this in a lot of different scenarios different color shirts really didn't seem to have a negative impact neither did setting up an edge or kicker light to you know identify your hair separately from your background that didn't really seem to matter at all even if my hair was super dark and i had a dark background it never cut into my head or anything like that so while it may still have artifacts of bringing other things into the frame unlike other options i never had artifacts of it actually removing parts of my body or head which is freaking impressive it also seemed to handle hats fine as well too headsets are kind of hit and miss as usual which is true of even cheap green screen setup so there's that it seemed to detect them fine and didn't really crop them out so much as the gaps between the headset and my head you know usually still showed the background more often than not desktop microphones seem to be a really big problem however i didn't have a ton of time to test this particularly out too much but just pulling my main microphone down in front of me it immediately keyed out the whole mic arm and the mic itself right in front of my face obviously that produces undesirable results even if i could have some weird fun with it what's cool is that the effects here work regardless of how sharp or in or out of focus your camera is since it's looking for the human face first since it's looking for the human shape first obviously it's gonna you know work better with more detail and more focus but even when my a7s ii's autofocus started acting up i didn't really have any issues obviously my image was blurry but in terms of background removal it was fine i also didn't see too significantly better of an improvement going between different quality webcams i had worse results with this cheapo webcam that only did 30fps at 4x3 aspect ratios according to nvidia broadcast but it wasn't that much worse here's a quick comparison between nvidia broadcast versus a typical mediocre twitch green screen setup while the chroma key removes things more effectively and it's not going to magically just put something else in your background it leaves a green outline and color cast on me which i just hate seeing in videos there's definitely a trade-off i also tested to see if as with the now defunct old personify chroma cam software that shipped with the c922 webcam if using a green screen would actually help the background removal software it kind of did because now you don't have a background you just have the green but then you still have the green cast and outline before we go any further let's talk about how this double round of content was possible for me this week i'm now part of nebula i've partnered with a bunch of creative friends for our own platform where we don't have to worry about demonetization or fighting an algorithm here i'll have all of my normal content that you know and love but also extra and exclusive content that goes beyond what i can show on youtube all ad free the platform combines some of the top education creators from youtube as well as unique collaborative opportunities that you might not normally see and while they were at it the people at curiositystream saw what was happening over at nebula and wanted to partner up to support even more educational content so now we have a deal with them where if you click the link in the description below you have access to curiosity stream and all of its great educational and documentary content as well as all of nebulous content from us education youtubers for free and curiosity stream is offering a 26 discount on their annual price which knocks it down to less than 15 for a year's worth of content and supporting educational creators like myself while you're there i recommend checking out great content such as alex goes bananas which is where alex nicholl discovers the pop culture i grew up with and makes me feel like i'm getting old because i am making this video that made me feel like i was getting old as well as another alex low-spec gamer's new podcast genesis which covers the origin stories of your favorite youtubers if you want to support my channel and support educational content on the whole check out curiositystream.com epos to sign up for just 15 per year that's better than most ways you can support channels so thanks while it's all cool and great there are some frustrating limitations with the software at this point in time i already mentioned the current state of device support so i won't talk about that too much more but there are a couple other points firstly you cannot use your webcam in nvidia broadcast and a streaming gap at the same time this is a windows limitation that means that there's no easy way to have different scenes where one has the background removed over top of your gameplay and then the other doesn't for like just chatting or something and there's also no hotkey support for disabling it either so you have to use the tray icon with your mouse which can be a problem in the middle of a stream and this is because not being able to use it in two places is because it occupies your webcam like any other program in windows when windows does not allow uvc devices to be shared through multiple apps at a time this will be a frustration for some secondly and probably most importantly imo you have no way to actually adjust your webcam when it runs through nvidia broadcast you can't bring up the uvc controls to adjust your brightness your gain your exposure your white balance and so on you can of course apply color correction to the nvidia broadcast camera and obs or whatever after the fact but you can't set any manual controls it kicks it back to auto so if your exposure is wrong on the auto settings you cannot fix it so most of my webcam samples just have me completely overexposed because i can't adjust that i tried setting it using the uvc controls in obs and then deactivating it and adding it to nvidia broadcast but that just kicked it all back to auto settings so there was no way to fix it this needs to be a high priority to fix with this software in my opinion next up we have performance numbers i discussed this a bit in my rtx 3080 review but i want to cover performance more first in terms of performance is stability again this is still in beta it's still being worked on and while i found it to be more stable than the initial runs of rtx voice in the first place i did have instances where the camera would go gray on me or just like disappear or deactivate itself and i'd need to restart the whole app or something else would happen and i would not find my devices at all until i completely rebooted my system little quirks while nvidia broadcast does not use the 3d game rendering part of your graphics card at least for the main workload it still does use cores on it to do the processing and that will impact your game while streaming with it enabled the severity of the impact will depend on your game however generally games in general will see a minor impact but newer directx 12 and ray tracing enabled games will see major impacts especially when using both voice and camera effects at the same time i did not test any with speaker enabled as well but i would imagine that adds just a little bit more extra load too i found that in some cases using both camera and mic had no extra impact over just the camera while in other cases it had a pretty huge impact it was kind of inconsistent i also ran into a weird issue with modern warfare 2019's warzone specifically which is an example of a dx12 rtx heavy game while enabling the camera effects and streaming in obs it just tanked my warzone performance to like 40 fps even in 1080p until i stopped and started streaming again and over in obs where it worse relaunched obs and got everything going again it was pretty inconsistent i did not include those results in my charts because it seemed to be a bug and i was able to recreate the same thing and get back up at 100 fps or whatever but it was really frustrating and happened a lot modern warfare 2019 is also a game where having both voice and camera on would definitely have an impact on how smooth and responsive the game felt compared to normal streaming so let's start with that game in terms of performance numbers at 4k i found the broadcast effects to have a 14 impact on game performance while streaming with reasonable settings on the 3080 over just streaming normally with a 23 impact on the 2080 over normal streaming so the 3080 with its extra rt cores can handle it better than the 2080 but there's still an impact on both and this will continue to be a trend throughout all of the games i test at 1080p 240 there was basically no impact uh on using it while streaming it was actually less of an impact than enabling max quality in ending settings on the 3080 but the 2080 suffered a 20 performance impact while using the broadcast software at 1440p enabling the broadcast effects on the 3080 saw an 18 impact over normal streaming but if you enabled rtx in the game since it uses similar cores i wanted to test that that bumped up to a 23 impact on your game in apex legends at 4k the 3080 had a 16 performance hit moving from normal streaming to streaming with broadcast while the 2080 had a 13 had just the 13 with the camera on but when you used camera and voice that went up to a 30 performance hit at 1080p 240 the rtx 3080 had only a 10 impact from the broadcast effects while the 2080 had a 19 from the camera effect and a 30 again from both of them being enabled so again the 2080 does not have the rt power to keep up here at 1440p 144 the 3080 only suffered a 22 percent impact from both broadcast effects last game here rainbow six siege at 4k 60 the 3080 saw a 15 impact with both effects whereas the 2080 only had a 12.5 percent impact from both but the frame rates that the two cards were running the game at are very different at 1080p 240 both the 3080 and the 2080 only i only had a six percent impact over normal streaming so they both tied there and then the at a 1440p 144 the 3080 only had an 8 impact with both effects that brings us to conclusion time overall this technology is phenomenal and i cannot wait to see where it goes from here because it can only get better however i really want them to also develop it as an after effects and resolve effect or plug in so that i can have more time per frame to process which should really produce better results i hate that all of this technology is so focused on just real-time performance that you know it gets you like 80 of the weight there in real time and so theoretically without doing it in real time by doing it in a video editor where it can have all the time in the world it should get you 100 of the way there but none of those tools ever end up or at least conveniently ending up in post-processing suites or software and so doing that by hand you know rotoscoping or whatever would take forever it would be health so i really hope this comes up as a video editor plug-in soon it works like magic but it's not actually magic so your results may vary and the impact on your streaming or on your game while streaming can be immense depending on the game like i wasn't even gonna try control with it because stoked to see future development from this and future training of course as i mentioned you can submit your own samples if everybody submits they have a lot more data points to work with and whatever but overall it's an incredibly valuable tool for streamers and content creators and i'm so glad that they're releasing for this for this i know i sound like a shield when i talk about it but the amount of you know time energy and development that nvidia has invested into streamers with software like this just cannot be overstated it's so freaking cool try it out for free give feedback and let me know what you think in the comment section down below hit the like button if you enjoyed go subscribe on nebula if you'd like or youtube for more tech education and stream content i'm your stream professor eposvox i'll see you next time i really need to stop talking now my ah i'm so glad this this crazy weekend is overall right today we're talking about the nvidia 30 series graphics cards again except not from a hardware perspective that video was yesterday or earlier today depending on when you watch this or when i post everything and how things are going today we're talking about the nvidia broadcast software this is software that they released alongside these graphics cards to filter out your voice for background sound or your chat member's voice for background sound as well as remove your background put you in front of any background or blur it out or whatever it's really cool software and i could not fit everything i wanted to cover in that one video so i'm making a second video about it let's go so this is nvidia broadcast this was formerly rtx broadcast but they're renaming it even though this specific software does actually require you to have an rtx card unlike the rtx voice that they officially open to all gtx cards as well this was shown at twitchcon 2019 and seemingly never heard from again until now it combines rtx voice with a virtual green screen app both built on lots of ai training models and provides some really powerful tools for streamers it still needs some work but it's a fantastic start and i can't wait to see where it goes and also i just can't wait to see when it gets added to video editors for post work because i think there's some huge potential there just to be clear this is not analyzing and sending back your video or audio to ai to then judge on the fly and change and learn as you go it's trained in advance and improves as more training can be done it was a weird myth that was getting spread when the original rtx voice launched i don't know what happened there this is of course a work in progress from a small team please give any feedback or contribute samples for training at broadcast.nvidia.com feedback all right let's walk through the software nvidia broadcast is this little app with three tabs two for voice and one for video let's start on voice formerly rtx voice though officially opened up to all nvidia cards with varying results this past week the audio portion of nvidia broadcast is designed to help filter out unwanted noise from your microphone or from voice chat if you have the hardware for that it has a quirk this is built on ai training of voices and microphones versus environmental sound to filter out just the voice in a way that at least in my previous round of testing i found to be way higher quality and more natural sounding than the results from most real-time noise suppression or noise removal options it's quite impressive as you can see here here's a vacuum the vacuum is running the vacuum is running the vacuum is running the vacuum is running the vacuum is running the vacuum is running the vacuum is running you'll always have some impact on the vocal sound due to frequencies matching between whatever background sound and your voice but cutting out a vacuum is crazy and yes i was the first one to do this on a youtube video everyone else seemed to just rip me off and then got all the credit oh well this does work best with cardio and microphones such as desktop streaming mics or headset mics it works less well on boom mics and even worse on omni-directional mics such as lav mics i will have more testing of this in the future i just haven't had time yet for microphone settings choose your microphone device as an audio source from the list toggle on the noise removal and set the strength of the filter they even provide a neat little tool to record and play back a clip of your voice to see how the filter affects it so you apply it normally listen to it then up you know increase the strength listen to it again until you get to a sweet spot that gets you what you want pretty straightforward that's it for speakers you do the same thing but with an audio output device this is for voice chats or conference calls if someone has a lot of background noise or keyboard sound or something really obnoxious they have simple background sounds available to playback as wave files so you can hear how the strength of the filter affects them to get you know a sweet spot of where you want so that is not too overbearing however a problem you may run into with the speakers is that unless you're purely doing a voice chat setup for your audio you don't have any other sound going on you'll need a dedicated output for your voice chat to run through because otherwise it'll filter out every other sound coming through your system your game sound your music your video you're watching anything so if you're using a physical hardware mixer or separate without separate output runs or something like voice meter software that's fine but if you have everything running to one device that will present a problem it's also a problem for the go xlr as the filter sound is actually presented as a separate output device that you then listen to not added back into the original output device that wouldn't work so then you can't add it to your go xlr the elgato wave microphones however do not have this problem because you can add devices to the software then you have the camera beta tab here you select your camera source the resolution and frame rate to run it at your effect and a preview of it happening currently it mostly supports uvc devices only so devices that present as webcams this is mostly limited to actual webcams but of course it also includes capture cards that present as uvc devices as well including the cam link the magewell usb cards the extreme cap uvc the hd60s plus as with the cloner alliance flint 4kp plus i reviewed as well as the 15 cant link that i made a video on and has since spread over the summer like crazy that's also uvc and so with those uvc capable capture cards you can use your real camera with hdmi like i am here in order to run it through nvidia broadcast so that's fine more expanded capture card support will be added in time and i really really really hope and i'm gonna hammer home that i really want them to add support proper support for black magic capture cards as these present differently than most capture cards and are usually left out but literally all of my camera outputs run through black magic cards and i really don't want to get left behind because they don't want to support it the resolution and frame rate that you choose impacts the workload of the neural network related cores on the graphics card in order to produce the image so if you're only doing a small face cam view in the corner of your stream you'll want to run it at a lower resolution so that it will have to work less and impact your game performance even less or have better performance for the processing overall i have not found lower resolutions to significantly impact the quality of the effect at all uh but there will be some differences by nature of there being less detail you know being resolved it does run at 4k though so that's pretty cool there's four effects to choose from for your camera back out background blur background replacement background removal and auto frame unfortunately auto frame cannot be run at the same time as other effects background blur is fairly self-explanatory it blurs out everything but you and gives you that faux bokeh depth of field look that i have right here this isn't faux though this is the real thing if i get close you can see it change that's what a wide aperture lens does it's especially convincing to use to enhance an already somewhat blurred out background from a lens from a mirrorless or dslr camera so if you have a real camera and it has some degree of background blur but not as much as you would like you can enhance that with this feature and it looks a lot more natural because any kind of imperfections are going to blend in a little bit more because it's already somewhat blurred out so that's pretty cool you all know about now that my office is also a living room where most of our baby furniture is since that's where he spends most of his time and a garage with my server rack and workbench behind it so whenever i need to record videos in that space background blur is actually really useful here there's a slider to control the intensity of the blur background replacement removes your background entirely and places an image behind you i use jensen's kitchen provided in the nvidia community pack linked below i recommend making sure that your image is decently blurred you can add that blur effect in photoshop or or critter or whatever as you like um as a super sharp image especially one that's more sharp than say your webcam is very distracting and much less appealing to look at background removal just removes your background and gives you a transparency compatible video that you can add to your streaming program and add your stream without using actual chroma key so why would you use background replacement over removal well for video calling and direct recording of video or vlogging scenarios where transparency won't be maintained the full removal gives you more control to manage in your streaming app but in a video call or direct recording all of that transparency is lost so you need to already have something presented behind you so that's where that comes in handy auto framing is pretty cool it will zoom in a bit and otherwise move your video feed around based on panning the shot around to keep yourself focusing in the center of the frame regardless of how much you move or shake or whatever you get a slider to control how much the initial zoom or crop factor is so that it has more room to work with within your frame so i recommend using a higher resolution image with that if possible um it's pretty cool some people like it unfortunately you can't do this and remove your background at the same time i realize that could look pretty silly but some people might want to you can't nvidia broadcast lives in your system tray and gives you a quick and easy right-click menu option to toggle camera speaker and microphone settings on or off in case they're impacting your game too much which i kept having having happen in modern warfare or if you have any issues or whatever you can quickly toggle them off next let's talk quality we're going to focus on the video side of things as i've already talked about voice in multiple videos at this point i'll have my samples linked in the video description overall the results from nvidia broadcasts for the background adjustment are pretty great depending on what's in your background of your actual shot the results can be beyond impressive when it's really working it feels revolutionary some of the clips where i'm only really lit by my computer monitors and using a crappy c930e webcam are so incredibly impressive it's been hard to keep keep this a secret until the embargo lifts but it's not perfect it does struggle with some of the same artifacting issues that most virtual background removal software do struggle with i have found darker backgrounds that work better than lighter ones as too many bright colors or shapes behind you can cause chunks of things to be detected along with you also having a second person behind you can confuse it a whole heck of a lot it's built for one streamer for example in terms of how it works i've been using it on my gaming side where i have much less going on in the background it's just our desk and the wall in the kitchen it's much further away that gets a lot better results especially with the lights off than my full living room background that i usually have with all the lights on as there's so many big shapes and colors and it just kind of confuses the the detection what's impressive is that when it is detecting right the edges of the cut around you look completely natural you can't even tell that there's there was ever something back behind you in the first place and you get to avoid the fuzzy green outline and color cast that a green or blue screen would cause however chairs are a problem the team had only a limited number of chairs behind users to train ai on and my primary chair the ergo chair 2 in gray confuses it even more there's no chair detection toggle like some options so you you know if the chair is always left in but if your chair does not look like what it expects you could have some issues it seems to prefer black chairs so swapping to my black respawn chair for the most part works better or if you have a mid-back chair like the one i'm sitting in right now it works even better because for the most part you're going to be blocking it with your body if your camera is angled down too far and it shows like your arm rests and stuff like that that could cause a lot of weird artifacting too because it has really like you can't expect it to train for that so you'll probably want to cut crop off the bottom of that feed a little bit so i tested this in a lot of different scenarios different color shirts really didn't seem to have a negative impact neither did setting up an edge or kicker light to you know identify your hair separately from your background that didn't really seem to matter at all even if my hair was super dark and i had a dark background it never cut into my head or anything like that so while it may still have artifacts of bringing other things into the frame unlike other options i never had artifacts of it actually removing parts of my body or head which is freaking impressive it also seemed to handle hats fine as well too headsets are kind of hit and miss as usual which is true of even cheap green screen setup so there's that it seemed to detect them fine and didn't really crop them out so much as the gaps between the headset and my head you know usually still showed the background more often than not desktop microphones seem to be a really big problem however i didn't have a ton of time to test this particularly out too much but just pulling my main microphone down in front of me it immediately keyed out the whole mic arm and the mic itself right in front of my face obviously that produces undesirable results even if i could have some weird fun with it what's cool is that the effects here work regardless of how sharp or in or out of focus your camera is since it's looking for the human face first since it's looking for the human shape first obviously it's gonna you know work better with more detail and more focus but even when my a7s ii's autofocus started acting up i didn't really have any issues obviously my image was blurry but in terms of background removal it was fine i also didn't see too significantly better of an improvement going between different quality webcams i had worse results with this cheapo webcam that only did 30fps at 4x3 aspect ratios according to nvidia broadcast but it wasn't that much worse here's a quick comparison between nvidia broadcast versus a typical mediocre twitch green screen setup while the chroma key removes things more effectively and it's not going to magically just put something else in your background it leaves a green outline and color cast on me which i just hate seeing in videos there's definitely a trade-off i also tested to see if as with the now defunct old personify chroma cam software that shipped with the c922 webcam if using a green screen would actually help the background removal software it kind of did because now you don't have a background you just have the green but then you still have the green cast and outline before we go any further let's talk about how this double round of content was possible for me this week i'm now part of nebula i've partnered with a bunch of creative friends for our own platform where we don't have to worry about demonetization or fighting an algorithm here i'll have all of my normal content that you know and love but also extra and exclusive content that goes beyond what i can show on youtube all ad free the platform combines some of the top education creators from youtube as well as unique collaborative opportunities that you might not normally see and while they were at it the people at curiositystream saw what was happening over at nebula and wanted to partner up to support even more educational content so now we have a deal with them where if you click the link in the description below you have access to curiosity stream and all of its great educational and documentary content as well as all of nebulous content from us education youtubers for free and curiosity stream is offering a 26 discount on their annual price which knocks it down to less than 15 for a year's worth of content and supporting educational creators like myself while you're there i recommend checking out great content such as alex goes bananas which is where alex nicholl discovers the pop culture i grew up with and makes me feel like i'm getting old because i am making this video that made me feel like i was getting old as well as another alex low-spec gamer's new podcast genesis which covers the origin stories of your favorite youtubers if you want to support my channel and support educational content on the whole check out curiositystream.com epos to sign up for just 15 per year that's better than most ways you can support channels so thanks while it's all cool and great there are some frustrating limitations with the software at this point in time i already mentioned the current state of device support so i won't talk about that too much more but there are a couple other points firstly you cannot use your webcam in nvidia broadcast and a streaming gap at the same time this is a windows limitation that means that there's no easy way to have different scenes where one has the background removed over top of your gameplay and then the other doesn't for like just chatting or something and there's also no hotkey support for disabling it either so you have to use the tray icon with your mouse which can be a problem in the middle of a stream and this is because not being able to use it in two places is because it occupies your webcam like any other program in windows when windows does not allow uvc devices to be shared through multiple apps at a time this will be a frustration for some secondly and probably most importantly imo you have no way to actually adjust your webcam when it runs through nvidia broadcast you can't bring up the uvc controls to adjust your brightness your gain your exposure your white balance and so on you can of course apply color correction to the nvidia broadcast camera and obs or whatever after the fact but you can't set any manual controls it kicks it back to auto so if your exposure is wrong on the auto settings you cannot fix it so most of my webcam samples just have me completely overexposed because i can't adjust that i tried setting it using the uvc controls in obs and then deactivating it and adding it to nvidia broadcast but that just kicked it all back to auto settings so there was no way to fix it this needs to be a high priority to fix with this software in my opinion next up we have performance numbers i discussed this a bit in my rtx 3080 review but i want to cover performance more first in terms of performance is stability again this is still in beta it's still being worked on and while i found it to be more stable than the initial runs of rtx voice in the first place i did have instances where the camera would go gray on me or just like disappear or deactivate itself and i'd need to restart the whole app or something else would happen and i would not find my devices at all until i completely rebooted my system little quirks while nvidia broadcast does not use the 3d game rendering part of your graphics card at least for the main workload it still does use cores on it to do the processing and that will impact your game while streaming with it enabled the severity of the impact will depend on your game however generally games in general will see a minor impact but newer directx 12 and ray tracing enabled games will see major impacts especially when using both voice and camera effects at the same time i did not test any with speaker enabled as well but i would imagine that adds just a little bit more extra load too i found that in some cases using both camera and mic had no extra impact over just the camera while in other cases it had a pretty huge impact it was kind of inconsistent i also ran into a weird issue with modern warfare 2019's warzone specifically which is an example of a dx12 rtx heavy game while enabling the camera effects and streaming in obs it just tanked my warzone performance to like 40 fps even in 1080p until i stopped and started streaming again and over in obs where it worse relaunched obs and got everything going again it was pretty inconsistent i did not include those results in my charts because it seemed to be a bug and i was able to recreate the same thing and get back up at 100 fps or whatever but it was really frustrating and happened a lot modern warfare 2019 is also a game where having both voice and camera on would definitely have an impact on how smooth and responsive the game felt compared to normal streaming so let's start with that game in terms of performance numbers at 4k i found the broadcast effects to have a 14 impact on game performance while streaming with reasonable settings on the 3080 over just streaming normally with a 23 impact on the 2080 over normal streaming so the 3080 with its extra rt cores can handle it better than the 2080 but there's still an impact on both and this will continue to be a trend throughout all of the games i test at 1080p 240 there was basically no impact uh on using it while streaming it was actually less of an impact than enabling max quality in ending settings on the 3080 but the 2080 suffered a 20 performance impact while using the broadcast software at 1440p enabling the broadcast effects on the 3080 saw an 18 impact over normal streaming but if you enabled rtx in the game since it uses similar cores i wanted to test that that bumped up to a 23 impact on your game in apex legends at 4k the 3080 had a 16 performance hit moving from normal streaming to streaming with broadcast while the 2080 had a 13 had just the 13 with the camera on but when you used camera and voice that went up to a 30 performance hit at 1080p 240 the rtx 3080 had only a 10 impact from the broadcast effects while the 2080 had a 19 from the camera effect and a 30 again from both of them being enabled so again the 2080 does not have the rt power to keep up here at 1440p 144 the 3080 only suffered a 22 percent impact from both broadcast effects last game here rainbow six siege at 4k 60 the 3080 saw a 15 impact with both effects whereas the 2080 only had a 12.5 percent impact from both but the frame rates that the two cards were running the game at are very different at 1080p 240 both the 3080 and the 2080 only i only had a six percent impact over normal streaming so they both tied there and then the at a 1440p 144 the 3080 only had an 8 impact with both effects that brings us to conclusion time overall this technology is phenomenal and i cannot wait to see where it goes from here because it can only get better however i really want them to also develop it as an after effects and resolve effect or plug in so that i can have more time per frame to process which should really produce better results i hate that all of this technology is so focused on just real-time performance that you know it gets you like 80 of the weight there in real time and so theoretically without doing it in real time by doing it in a video editor where it can have all the time in the world it should get you 100 of the way there but none of those tools ever end up or at least conveniently ending up in post-processing suites or software and so doing that by hand you know rotoscoping or whatever would take forever it would be health so i really hope this comes up as a video editor plug-in soon it works like magic but it's not actually magic so your results may vary and the impact on your streaming or on your game while streaming can be immense depending on the game like i wasn't even gonna try control with it because stoked to see future development from this and future training of course as i mentioned you can submit your own samples if everybody submits they have a lot more data points to work with and whatever but overall it's an incredibly valuable tool for streamers and content creators and i'm so glad that they're releasing for this for this i know i sound like a shield when i talk about it but the amount of you know time energy and development that nvidia has invested into streamers with software like this just cannot be overstated it's so freaking cool try it out for free give feedback and let me know what you think in the comment section down below hit the like button if you enjoyed go subscribe on nebula if you'd like or youtube for more tech education and stream content i'm your stream professor eposvox i'll see you next time i really need to stop talking now my ah i'm so glad this this crazy weekend is over\n"