Question / Help Streaming with overlays whilst recording without them?

[Duelsy]

Hi,

I’m a relatively new content creator who streams and creates YouTube videos. I currentlyuse a single PC setup. I've been using Streamlabs for the last 4 months because it has features like “selective recording”, allowing me to stream with a web cam and alerts etc but record without them showing i.e. "clean footage". However when using the NVENC (new) their software doesn't allow me to rescale for my stream, so I have to scale down in my video settings (rather than my output). This means I get sub-par footage which I can’t use. Therefore I'm looking to move over to OBS, as you can set separate resolutions for everything. I've been messing around with stuff today and OBS seems to run better whilst consuming less resources which is good.

Is there a way in OBS to stream with overlays but record a clean scene without them for YouTube (i.e. just the game capture)? If there isn't an option for this what's the best solution, simply to run two instances of OBS?

Just for additional info - I've been using NVENC (new) for both recording and streaming. I notice my GPU shows 100% usage in task task manager when doing this, although I don't get any skipped frames etc. I've also been experimenting with recording using NVENC (new) whilst streaming using x264 on fast or medium. In my setup I personally believe my CPU is going to be the bottleneck.

PC specs below so everyone knows what I’m working with. Also please note I run OBS in admin mode.

PC specs:
GPU - RTX 2080Ti
CPU - i7-8700K running at 5.1Ghz
RAM - 32GB Corsair Dominator
Gaming monitor resolution - 1440P

If you need any more info or some screenshots etc please let me know.

Thanks for your help.

 

[FerretBomb]

Not at present. A proper implementation of selective recording (as opposed to the patch-job Streamlabs is using) is planned, but has no ETA at the moment. One workaround if all you are looking for is clean gameplay footage, you can try installing the obs-virtualcam plugin, and add a VC as a Filter to your gameplay capture source. Then run a second instance of OBS in Portable mode, just recording that 'webcam'.

NVENC on the 2080 is equivalent to x264 Slow.
NVENC is also a separate part of the GPU die that normally sits idle, doing nothing. Encoding with NVENC will not affect your gameplay, unless you use CUDA-enabled options (Max Quality preset, Psycho-visual Tuning, and Lookahead; just use Quality instead, and turn the other two off).
NVENC on consumer-grade nVidia cards also supports TWO encoding streams simultaneously, up to 4K@60fps without issue. You MAY have to disable Shadowplay and Windows' GameDVR features to 'free up' the second encoding slot, but you can absolutely use both... one for recording, one for streaming.

If anything, I'd recommend using x264 software for your local full-overlay recording with CQP or CRF (quality target) encoding, and a fast+ preset... it'll minimize CPU usage since local bandwidth writing to a HDD is cheap and plentiful, without the bitrate bottleneck of streaming. Use NVENC on that card for the stream, it'll give you better quality. Then use the other NVENC session for the 'webcam' recording, to help keep load minimal and quality high (again, using CQP/CRF).

 

[Duelsy]

Yes all I want from the recording is clean gameplay footage. I can then use it in a "movie" format or commentate over it for YT. I'm still recording my voice anyway, so could lay it over during editing if required.

Hmm okay interesting, thanks for the info. The tutorials I'd watched regarding NVENC new all said to use max quality, physcho-tuning etc etc but none of them actually explained anything about CUDA options or what that entailed. Will definitely do some more reading about that tomorrow morning. Is there a way to check whether one of the NVENC encoder slots is already in use by another programme?

 

[FerretBomb]

Yes all I want from the recording is clean gameplay footage. I can then use it in a "movie" format or commentate over it for YT. I'm still recording my voice anyway, so could lay it over during editing if required.

Hmm okay interesting, thanks for the info. The tutorials I'd watched regarding NVENC new all said to use max quality, physcho-tuning etc etc but none of them actually explained anything about CUDA options or what that entailed. Will definitely do some more reading about that tomorrow morning. Is there a way to check whether one of the NVENC encoder slots is already in use by another programme?

99.999% of Youtube 'best settings guides' are complete garbage; just information being parroted by people who don't understand it or the underlying mechanics. nVidia does recommend enabling PVT, but it doesn't really help much, and tends to result in encoding overload a LOT of the time. It's really best to just leave it off and not have the problems it regularly causes. Max Quality is the same as Quality, but with 2-pass mode enabled which isn't going to work well during live recording... it's fine for transcoding though.

Unfortunately there isn't; you just try using it, and if it errors out, it's in-use. Disabling Shadowplay and GameDVR normally gets it freed up though.