[HELP] Recommended OBS Settings for Youtube?

[Librewolf]

Recommended OBS Settings for Youtube?

I've just updated from OBS 26.1.1 to 29.1.3 and it seems alot has changed...
To be honest i don't know to much about video recording software and all it's settings.
I am sure i'm not the only person asking these type of questions here so feel free to redirect me to whatever might anser my questions, thank you.
That said, i plan on recording desktop videos, various types of things that could be
OS configuration or the webbrowser or videogames, to upload it on google's censorship sh!thole we all know: youtube.
So my questions...

• Recording format:
Matroska Video (.mkv) this is default, i don't know if this can be uploded to yt?
MPEG-4 (.mp4) i believe this should be used for youtube?
Fragmented (.mp4) i don't know what this is.
The other recording format options seemed irrelevant, correct me if wrong.

• Video encoder:
use stream encoder is the default, dont know what this means.
i would select nvidia nvenc H.264 asuming HEVC is older? (yeah i've got an nvidia gpu)

• Audio encoder
I'm lost here, no idea what i'm doing.
I have a soundcard (coundblaster AE7) which serves me well.
If i want the best possible audio quality what do i select?
ALAC 24bit, FLAC 16bit, AAC, OPUS, stream encoder?

• Rate control:
i've always used CBR back the days, but now, CQB, VBR, for best quality what do i chose?
lossless needs to much storage.

• Bitrate:
so far i remember youtube supports only max 25.000 Kbps so recording higher than that is pointless?

• Other questions

What is the multipass mode, can i ignore this settings if i dont intend to use it?

What does the max B-frames setting do and what do you recommend to select here?

Would you say 320 audiobitrate is over the top?
I noticed it increases the filesize quiet a bit.

In advanced settings theres only Direct3D 11 renderer available, this must be DX11.
Are there any news on vulkan or dx12 or opengl?

Also in advanced settings, what color format should i use?
I tend to not change settings when i don't know what they do but i also want to learn
what this is.
As for color space, Rec709 is the default. Is that superior to sRGB?
So far i remember sRGB is what my monitor's profile is.

When i add a new capture source, what's the difference between display and game capture?
Obviously one is the entire screen the other a specific program.
I mean, does it impact performance? Does it matter?

 

[Harold]

The most straightforward way to record:
Output mode: Simple
Recording quality: Indistinguishable
Recording encoder: Nvenc h.264
Recording format: mkv

If you are not editing, youtube is able to accept mkv recordings without additional processing. If you are editing, some editors require you to use the remux recording function in obs' file menu in order to be able to use them.

color format, space and range: NV12, 709, Limited

The color format/space/range is that because the entire professional video industry has standardized on the combination.

In advanced settings theres only Direct3D 11 renderer available, this must be DX11.
Are there any news on vulkan or dx12 or opengl?

There's no functionality or performance benefit to using DX12 over DX11
OpenGL used to be selectable, but it was unreliable, unstable and under performing.

When i add a new capture source, what's the difference between display and game capture?
Obviously one is the entire screen the other a specific program.
I mean, does it impact performance? Does it matter?

Game capture uses a form of texture sharing that allows the uncompressed frames to be captured from the game and sent to scene composition without directly involving the CPU at all, greatly improving performance. display capture gets the frames in such a way that they have to be sent over the pci-e interface of the video card.

 

[Librewolf]

Thanks but i still have questions.

• Audio encoder
In the soundblaster software i can select 32bit 96kHz for my headphones.
In audacity i can select project rate (Hz) there's alot of options but i always select 96000 since that's equal with the soundblaster software.
In obs studio i can select the audio encoder (i would go for ALAC or FLAC here) and i can also change the audio bitrate to at max 320.
So while the soundblaster software and audacity seemingly (or i'm blind) have only one audio quality related option (kHz / Hz), obs studio has two.
From my current knownledge i would use FLAC-16bit with 320 bitrate.
Why is it that FLAC uses 16 bit and ALAC uses 24 bit, theoreticaly that means using FLAC has worse quality than ALAC?
From what i've read online, both are losless anyway?

 

[Harold]

OBS' default audio encoder is AAC, not FLAC or ALAC, and audio quality in actual perception terms is not significantly different in the increments above 128-160kbit. There's not a lot of quality gain going above that range, and there's even less gain if you have itunes installed so that coreaudio is used for AAC encoding instead of ffmpeg.

using FLAC or ALAC as your audio encoder will cause it to more generally ignore your chosen bitrate because of their lossless nature.

FLAC seems to have better compatibility in editing software and media players than ALAC as well.

 

[Librewolf]

Thanks again for the explanation.

May i still ask a few more questions..
You said the most straight forward way to recording is using recording quality: Indistinguishable
I tried that once and it significantly raised the filesize, so uploading 30 minute long videos that have a crazy file size in GB's takes a long time on youtube, for that reason i would choose the next best quality option, what would you recommend there?
I've used CBR 30000 Kbps so far, and i know nothing about the other options CQP and VBR.
In another obs forums thread, i've read that CQP is a good option, how does it compare to CBR?

I do intend to edit videos (not always but still) so what's the next choice after mkv? mp4?
If i remember correctly one of the first videos i tried to upload on youtube was using the mkv format and youtube said i can only upload mp4 videos so that wasn't even possible, atleast back then.

Should i leave the keyframe-interval on default?

You said that color format, space and range: NV12, 709, Limited is the video industry standard.
That in mind, i've just looked through the color format options and there's some 8bit, 10bit and 16bit available.
Wouldn't it make sense for the best quality to select a higher value?
My monitor specs:
color depth: 8bit
color format: RGB
color range: standard dynamic range (SDR)

 

[Harold]

CQP is a quality based rate control, it will always target the correct bitrate for the targeted quality level.
Lower motion, lower picture detail content will have smaller file sizes than high motion high picture detail.

I do intend to edit videos (not always but still) so what's the next choice after mkv? mp4?

NEVER save directly to regular mp4. If it does not finalize properly, the ENTIRE recording will be lost.
Fragmented MP4 does attempt to mitigate the issue, but it has poor compatibility in editors.
Saving to MKV then remuxing to mp4 specifically if the editor requires it is the safest option.
Blackmagic Design's Davinci Resolve and Resolve Studio are able to load MKV files without remuxing.

If i remember correctly one of the first videos i tried to upload on youtube was using the mkv format and youtube said i can only upload mp4 videos so that wasn't even possible, atleast back then.

Youtube has been able to accept mkv format for over ten years at this point, and it may be closer to 20 years, given the age of the format.

color range: standard dynamic range (SDR)

Because of this on its own, the HDR color format options are a waste. Your system doesn't present you with them. No point in wasting space on things your system won't use.

 

[Librewolf]

Thank you.

I don't understand what you ment by "NEVER save directly to regular mp4. If it does not finalize properly,"
I've used mp4 as recording format all the time, hit a keybind to start recording and hit another keybind to stop recording, then opened the recording with vlc mediaplayer and it was fine, also editing in adobe premiere worked fine, what exactly do you mean by "finalize properly"?
It's strange that you say mkv was accepted by youtube for such a long time, i remember clearly that i tried uploading a .mkv video to youtube, then there was a prompt saying this file format is not accepted/supported. I will try again.

Alright i will use the standard color settings as you suggested, thanks again.

 

[Harold]

MP4, when you interrupt the recording process, functionally deletes the entire recording.
"Stop recording" is not an interruption. It's a graceful stop.
if OBS crashes mid-recording, you lose power, you bluescreen, or anything like that, the file is unusable if you save directly to mp4.

 

[Librewolf]

Right that's problematic for some people, i haven't had bluescreens, crashes or powerloss in a long time so mp4 should be fine but its certainly good to be aware of.

Quick update, i just did a 20 minute recording of resident evil 4 remake with the highest quality settings (losless).
The result is shocking... 60 GB. Seriously 60 GB for 20 minutes of recording, this is insane.
What's the best config for high quality recording while maintaining a decent storage size?

 

[Harold]

That's why I said indistinguishable rather than lossless.

 

[Librewolf]

That's why I said indistinguishable rather than lossless.

Oh, my bad.
Where we talking about the same thing, > Settings > Output > Ratecontrol ?
There is no "indistinguishable" option there, what i see are those:
CBR, CQP, VBR, Lossless.
You ment one of these is indistinguishable?

I want to get back to one of your previous replys, you said that game capture barely uses any cpu resources, so i asume this capture mode is more gpu intense?
You also said that display capture get's the frames sent over the pci-e interface of the gpu.
Does that not happen with the game caputre mode?
How exacly is the cpu more involved via display capture and less with game caputre?
I don't understand, if the game caputre mode greatly improves performance by using less cpu resources, or using the gpu more efficent, why is does this only work with capturing games or other fullscreen applications, can the same technique not be used on the whole screen?

If you may let me get back to another reply you sent,
"128-160kbit. There's not a lot of quality gain going above that range"
I've had crap headphones for few years, then not to long ago got the soundblaster AE7 (soundcard) and sennhaiser hd 600 headphones.
Seriously couldn't believe how good the audio quality was when i first had that setup, even bad audio sounded much much better.
I did listen to mp3 and FLAC audio, i could swear that FLAC sounds better.
PS: I never used itunes, i hate apple. BS company.

Now replying to:
"Using FLAC or ALAC as your audio encoder will cause it to more generally ignore your chosen bitrate because of their lossless nature."
Can you explain what you ment by generally? If i use FLAC, does it matter if i use 64 or 320 audio bitrate, if yes why isn't the setting greyed out basically telling the user "this is controlled by FLAC"?

I was just editing a video in adobe premiere pro, that video was recorded with FLAC and guess what, adobe doesn't support it.
Also doesn't support ALAC, only the AAC standard.
So i'm screwed, either find a plugin for adobe to support FLAC or revert back to AAC.
What i don't understand, why doesn't AAC highligh what bitrate it uses? FLAC clearly notes it uses 16bit and ALAC 24bit.

 

[Librewolf]

Oh, my bad.
Where we talking about the same thing, > Settings > Output > Ratecontrol ?
There is no "indistinguishable" option there, what i see are those:
CBR, CQP, VBR, Lossless.
You ment one of these is indistinguishable?
I want to get back to one of your previous replys, you said that game capture barely uses any cpu resources, so i asume this capture mode is more gpu intense?
You also said that display capture get's the frames sent over the pci-e interface of the gpu.
Does that not happen with the game caputre mode?
How exacly is the cpu more involved via display capture and less with game caputre?
I don't understand, if the game caputre mode greatly improves performance by using less cpu resources, or using the gpu more efficent, why is does this only work with capturing games or other fullscreen applications, can the same technique not be used on the whole screen?
If you may let me get back to another reply you sent,
"128-160kbit. There's not a lot of quality gain going above that range"
I've had crap headphones for few years, then not to long ago got the soundblaster AE7 (soundcard) and sennhaiser hd 600 headphones.
Seriously couldn't believe how good the audio quality was when i first had that setup, even bad audio sounded much much better.
I did listen to mp3 and FLAC audio, i could swear that FLAC sounds better.
PS: I never used itunes, i hate apple. BS company.
Now replying to:
"Using FLAC or ALAC as your audio encoder will cause it to more generally ignore your chosen bitrate because of their lossless nature."
Can you explain what you ment by generally? If i use FLAC, does it matter if i use 64 or 320 audio bitrate, if yes why isn't the setting greyed out basically telling the user "this is controlled by FLAC"?
I was just editing a video in adobe premiere pro, that video was recorded with FLAC and guess what, adobe doesn't support it.
Also doesn't support ALAC, only the AAC standard.
So i'm screwed, either find a plugin for adobe to support FLAC or revert back to AAC.
What i don't understand, why doesn't AAC highligh what bitrate it uses? FLAC clearly notes it uses 16bit and ALAC 24bit.

 

[koala]

Where we talking about the same thing, > Settings > Output > Ratecontrol ?
There is no "indistinguishable" option there, what i see are those:
CBR, CQP, VBR, Lossless.
You ment one of these is indistinguishable?

Harold recommended to switch to simple output mode, so "indistinguishable" becomes available, and with this mode, you're relieved from all the tiny settings you don't understand. You simply set the quality you want. Inside, simple mode is a wrapper around advanced mode and applies best practice, so it will usually produce the best result.

To explain what difference between bitrate rate controls (CBR, VBR) and quality rate controls (CQP, CRF) exist:

Inside, the encoder removes detail from video frames to achieve compression. If you ask the encoder to limit to some bitrate (CBR, VBR), it removes as much detail as required, until the compressed frames are so small to fit into the given bitrate.

With high motion video, where compression is difficult due to the massive changing between frames, extremely much detail is removed to achieve the given bitrate. So CBR/VBR is bad for high motion video, because the visual quality suffers for high motion. With low motion video, compression is very good due to not much changing between frames. So CBR/VBR is bad for low motion video, because the video will be blown up to the given bitrate, even if the changes could be compressed to less than the given bitrate.

The other rate controls (CQP, CRF) simply remove the same amount of detail from every frame, no matter the motion. This is defined with the CQ parameter (0..50). 0 removes nothing, so it is lossless, 50 removes so much everything is just a bloody mess. Sane values are 15 (indistinguishable from the original, but huge file size) to 30 (visible compression artifacts, but small file size). The encode outputs a variable bitrate according to the requested amount of detail to remove and related to the current motion. It outputs just enough to properly achieve the given quality, not too much and not too less.

Streaming has the constraint that not much bitrate is available, and the bitrate must be constant due to the inner working of network distribution. Due to this, CBR/VBR rate controls exist. For recording, disk space (the equivalent of bitrate) is unlimited, so use quality based rate controls such as CQP/CRF, since they don't squeeze or bloat the video into the forced bitrate.

 

[Librewolf]

Thanks koala, appriciate the valuable information!

From those two quality based rate controls (CQP/CRF) which one do you chose?
As for CQ parameter, what number is your take for the best quality while maintaining low file size?
Quote: sane values are 15 to 30, the only reason i bring this up again is because you mentioned visible compression artifacts.
Wondering what number you'd take to eliminate said artifacts (or minimize them as much as possible) but keeping file size is mind.
Where is that fine line between quality and filesize that professionals would recommend?
Also, how would i visualy notice these compression artifacts?
I know what artifacts are, correct me if wrong.. an artifacts is when there are left over's from the previous frame in the new frame.
But compression artifacts i can't make sense of.

According to you, CBR and VBR are using a fixed bitrate that doesn't adjust to low or high motion scenes and are for that reason bad for high motion capture, which makes sense.
But, couldn't i just rise the bitrate to such a value where the quality looks good even for high motion scenes?
I've always used CBR with 30000Kbps and thought it looks alright, but could be better.
Ofcourse increasing that number will also increase filesize.
You got me curious about non-fixed bitrates, that sounds smart.
How does that work, is there an algorythm detecting the amount of motion and adjusting the bitrate accordingly?
If this should be the cast, isn't there a delay with the algorythm detecting the motion causing the bitrate to be slightly "inacurrate"?

I will check out the simple-mode again, thanks for the headsup.

 

[Librewolf]

I've just made an 8 minute recording of resident evil 4 remake with the above settings.
Filesize: 5.81 GB
That's a bit to heavy. Should i lower the bitrate or change the recording quality? Or perhaps lower the encoder preset? Which of those?
In the simple interface i can't see any ratecontrol setting, what is it?
Is there any way to use FLAC with the simple interface? Sadly the audio encoder is locked with AAC in the simple mode, i tought you might know a tweak to bypass that, perhaps in a config file?

This will sound dumb... in the advanced interface, FLAC uses 16bit, ALAC uses 24bit and AAC doesn't say what bitrate it uses.
I can ofcourse set the bitrate to 320. And that's what i don't understand. 320bit is way higher than 16 or 24.
Is it the nature of how AAC works that it's less filesize and less quality while FLAC and ALAC have superior quality with higher filesize no matter what bitrate is set, or am i completely missing the logic?

 

[koala]

I don't choose any CQ value, I use simple output mode with "indistinguishable" quality myself. If you're curious what OBS chose internally, do a recording in simple mode and look afterwards into the logfile what has been used for that session. OBS will record the exact encoder settings used.

The higher the CQ parameter, the more detail is removed. From 0 to 15, possibly 18, the human eye isn't able to detect any difference to the raw original. From 18 to 25, you're beginning to see somewhat washed and blurred areas. Above 23, these areas become definitely visible.
An increase by 3 about doubles file size, at least for the CQP rate control of Nvidia's nvenc.

If you're interested in how the CQ parameter changes the video, just capture an original 30 second video in lossless mode with low motion and high motion parts. Then change to advanced output mode and add a media source to play back that lossless video instead of capturing a new video. Then choose CQP with 5, then 10, than 15, then 20, then 25, and so on up to 50, and record the playback again with these settings. Then compare the created videos in visual quality and file size.

OBS in simple output mode and indistinguishable quality uses a CQ of about 17-20, as far as I remember. It varies slightly with the resolution, because the higher the resolution, the higher the quality parameter can be without visual differences (or was it the other way round? Don't remember. Because of this, I'm using simple mode to let OBS take care of this).

You can of course use CBR and choose an arbitrary high bitrate, however this will waste disk space for low motion scenes and is still not enough for high motion scenes. A recording of 2560x1440 of some current computer game with indistinguishable quality will have about 120000 kbit/s. If this is too much for you, just choose a lower quality.

The algorithm you're asking is called rate control. It's bitrate control. For CBR, the encoder is starting with some very low quality first frame. If the resulting bitrate of the frames encoded in the last few seconds is lower than the requested bitrate, it increases the quality slightly. If the resulting bitrate of the so far encoded frame(s) is higher than the requested bitrate, it decreases the quality slightly. This is a in general the algorithm, slightly simplified. For CQP, the resulting bitrate is not checked. It's just always encoding with the CQ value the user gives to the encoder.

If it comes to audio, well, I never cared about audio quality for computer games, and I only record computer games. I always used the default AAC with the default audio bitrate.

By the way, you're confusing the bitrate (128 kbit/s) with the depth (16 or 24 bit). The depth defines how many numbers are available to represent one audio sample. A bitrate above 128 kbit/s and a depth above 16 bit isn't perceivable by the human ear of an adult. Only the ears of a child or growing up person might be able to hear any difference, however these people don't have "educated" ears usually, so they might hear a difference, but might not be able to classify what's different.

Higher bitrates and depths are used if the recording is made for postprocessing. Since any postprocessing takes a toll on quality and increases errors between analog original and digital sample, professionals want to minimize these errors, so they use better quality than humanly perceivable, so even the postprocessed material has still a quality high enough.

Wanting to record lossless audio might be an idea if you intend to record an audio session you play life with some instrument of your own, and intend to conserve and publish this in studio quality. However, if you just intend to capture some stuff from your browser or video game, this is just wasted.

 

[Librewolf]

I would lie if i said that this wasn't a little complicated, but i'm very thankful people here take the time to explains things properly evne when i have to read everything more than twice to understand it..

If it's not asked to much, would you mind letting me know your exact recording settings, that could be in textformat or a screenshot like the one i sent?
Just realized that my question whether to change the bitrate or the recording quality or the encoder preset didn't make any sense since i was reffering to the simple output-mode wheere the videobitrate and the encoder-preset are both options for streaming while only the recording quality is an option for recording... no idea how i could overlook that.
That means my choices with the simple output mode to limit filesize are way more limited than i thought, due to less configuration options it'll be difficult to have great quality with low filesize.
As i said, the 8 minute recording of resident evil 4 remake had a filesize of 5.81GB.
So to round that up, a 10min recording would have around 8GB, 20min 16GB, etc.
That's to much filesize in my opinion, mainly because the uploads will take ages on youtube.
Infact there are only two options i can think of that would lower the filesize with simple output mode.
These are, either to change the recording quality to something lower than indistinguishable, or under the video cathegory to set the output resolution to something lower than native.
My native resolution is WQHD (2560x1440p) so i would set the output resolution to HD (1920x1080) combined with the downscale-filter Lanczos-sharpened-scaling 36 samples. Or eventually bicubic 16 samples.
I don't really know what to do, with the current indistinguishable format the filesize is a bit to high. Don't you have the same problem with very long youtube uploads even for short videos becaue of the filesize? Anyways, any hints on what i could do to reduce filesize would be welcome.
I'll admit indistinguishable looks damn good and for that amount of quality the filesize seems low, but it's still a bit to much.

Good to know that bit-depth and bitrate (kbps) aren't the same thing...
So, what is the bit-depth of AAC? I could probably as you suggested make a recording and find out in the log files:
I'm just wondering why FLAC and ALAC show their bit-depth but AAC doesn't?

 

[qhobbes]

Settings > Output > Change Output Mode to Advanced > Recording tab > Video Encoder: NVIDIA NVENC H.264 (if your card supports HEVC and/or AV1, use those as better quality at bitrate ratio (unless you need to edit and you're editing software doesn't support those)). Here you can choose a Rate Control of CBR, CQP, VBR (variable bitrate, similar to CQP but with limits on bitrate) and lossless and other options.

 

[Librewolf]

Hey, thanks for the reply.

As for video encoder i see the follow options:
use stream encoder, SVT-AV1, AOM AV1, NVIDIA NVENC H.264, NVIDIA HEVC, x264.
Does obs studio show these options to everyone or can i only see those that are compatible with my hardware?
My hardware: CPU i7-8700k GPU GTX 1080Ti, should support HEVC from the info i just looked up, can you confirm that?
In case my hardware doesn't support that, what would happen if i still set HEVC as encoder, not work at all, crashes, etc, i mean how would i notice?

There's this CPU benchmark software called OCCT, it has an option named instructions-se where it shows me AVX and AVX2, curious if this has anything to do with the AV used in obs studio?

 

[qhobbes]

"Does obs studio show these options to everyone" For the NVENC ones, I believe so (for people with any form of NVENC) because I have a GTX 660M and it shows those same options. It can only do NVENC H.264. Don't know about the AV1s

It supports it per https://developer.nvidia.com/video-encode-and-decode-gpu-support-matrix-new

AVX and AVX2 are instruction set extensions which can increase performance when the same operations are performed on multiple data objects. These can include SSE (Streaming SIMD Extensions) and AVX (Advanced Vector Extensions). OBS requires i5-2500K or better (or Ryzen 1300X which came out in 2017 with AVX2 or better) which has AVX and was released in 2011 along with AVX. x264, x265 and VTM video encoders can use AVX2 or AVX-512 to speed up encoding. Does it mean OBS requires that extension? I don't know, but as the YouTubers say, I digress.

Just use NVENC HEVC, better quality for the bitrate.