Best Quality Settings OBS

[Shwen]

Hi!

I'm looking to start a gaming YouTube channel, but most of my captures have a little washed out effect over them, but when I use DXTory, it looks almos exactly like the game. Graphics processing or hard drive space isn't a worry, I'm just looking for the best settings.

Specs: 1080, 6700K, 24GB RAM, 1TB HDD.

Thanks!

[NinJake]

Hmm... Since Shwen have you wanted to start a YT channel?

 

Good luck! It's a tough crowd these days with tons of people doing it. For OBS you'll just have to tweak your settings until that "washed out effect" goes away. This comes down to what types of games you're playing and how fast moving they are as well.

[Electronics Wizardy]

Just crank up the bitrate. 

[mariushm]

The washing out is probably because a video is marked as Full Range (YCbCr 0..255 for all 3) and the other is marked as Limited Range ( 16..235 for all, everything outside being considered invalid) but it's played on Youtube as full range. It's an issue of marking the h264 stream correctly (some flags in the file you upload to Youtube) , and also using the correct  matrix when doing the color conversions (for example using bt709 for HD content or bt.601/bt470 for SD)

 

In OBS Studio, you go in File > Settings > Advanced and you have there video renderer , color format , yuv color space and yuv color range.

 

Make sure color range is Full ,  Color space 709 is you stream at anything higher than 1024x576 (most video players will consider that as DVD 16:9 and DVDs were encoded with bt601 color space so if you compress such resolutions using bt709, the difference in color spaces could make some colors look different on some players which make the assumption the stream was encoded as bt601).

 

For Color Format, whatever works better with your video cards. The most "friendly" to the video compressor and the video cards are NV12 followed by I420 ... RGB should only be used if your streams have a lot of  2D / screen capture.

 

x264 software encoder works natively with YV12, which is a slightly different version of NV12 , basically there can be conversion done between these two formats without any loss of quality. Hardware h264 encoders should work with NV12 or YV12 just fine.

 

 

[Omon_Ra]

For just recording, I use these settings. Look at the lower 'Recording' box, ignore the 'Streaming' one. You may want an SSD for your recording drive.

 

This is an example of the quality.

 

 

[mariushm]

Yeah, I'd do it the other way...

I'd use NVENC to stream on twitch using higher bitrate (Twitch allows up to 5-6 mbps these days)

 

Then I'd use x264 software encoder to save a higher quality to disk, which I could edit/recompress in a Youtube friendly format if needed.

For example, you could use something like

* bitrate up to 60000 (60 mbps)  - don't use cpu to squeeze as much quality in a specific amount, just use disk space

* CRF 5-8  - the general consensus is that below 5 you don't notice any difference visually, it's like saving a picture with JPG 99% quality, when your eyes usually only notice quality losses when you save JPG at 95% or less

* preset ultrafast - don't think too much, just quickly figure out some detail that can be removed to compress a second of video to use less than 60 mbps instead of 200-300 mbps

* colorprim bt709

* colormatrix bt709

* transfer bt709

 

You'd get a large video with variable bitrate up to 60mbps or whatever you allow, and compress with minimal cpu usage

Later, you can recompress to 20-25 mbps VBR with high quality settings to retain as much quality in the bitrate you have, or you can just upload that big file to youtube, because Youtube will recompress the video anyway.. so the more quality you give to youtube, the better your video would look.

 

[NinJake]

So I think what @mariushm is saying is to use nearly the highest settings you can for recording and let youtube deal with compression.

[mariushm]

Yeah, Youtube will recompress everything so the more quality you give it, the better your video would look.

 

You get more quality either by squeezing as much quality in a small bitrate (which means a lot of cpu usage in order to decide what can be kept and what can be left out, and more for better compression, which uses processor)  OR you just let the encoder use more disk space instead. Think of it like the difference between compressing files with FAST preset or Ultra preset in 7zip (with the ultra preset you get smaller file sizes but you also use more cpu)

 

The 6700k is a fast processor, but it's not quite super fast or super powerful, I'd think your interest would be to give as much CPU to the game and not keep cpu cores busy compressing video and potentially causing the game to produce less frames, or more inconsistent (variation in fps from second to second).

 

You can always recompress later the "quick and dumb compressed" large bitrate video to throw away even more detail that human eyes won't easily notice as gone, or to find content that's repeated in consecutive video frames and compress that repeated content better - at ultrafast/veryfast presets, in order to reduce cpu usage, such analysis is not performed or let Youtube do it for you.

 

[Mamonos]

If I understood correctly you talk about recording and in this case I would go for NVENC with CQP 14, high quality low latency preset.