Are these huge file sizes or are they normal?

[MrRandizzle]

I've been tinkering around A LOT with settings and trying to figure out how I can record 2 1080p60 videos at one time in OBS. I've gotten to this point with my settings with good quality and minimal dropped frames (<0.5%), but I'm finding that the file sizes are HUUUUUGE! This particular log file is associated with a 1 minute video that turned out to be 1.18GB! Unless my math is super bad, thats a bitrate of something like 160,000kbps!! [1.18GB/60s = ~0.2GB/s = ~160,000kbps]. I'm using CQP set to 16, which I've seen many people suggest in these forums, but I also read that the target bitrate for 1080p60 is anything from 12 to 50 mbps. So if I wanted to post a 30 minute video with gameplay AND a face cam using these settings, I'm looking at a 70.8GB video (before editing)? I feel like there's something wrong here. Is this unusual, or is this just what it takes to get a high quality 1080p60 video?

 

[carlmmii]

Any time you record using a quality-based target, it's going to depend on the complexity of the source video. Some content is easier to encode, some content takes more. This will have an effect on how big the overall filesize is given a set CQP value, but generally the more important factor is in the actual quality level you're looking for.

The recommendation of 14-16 for CQP is usually for editing-level quality, but depending on what you're actually using the recording for, this may not be wanted. Use your own eyes and do your own testing -- try higher CQP values, check the quality yourself, and adjust accordingly until you have the quality level that you actually want, or at least something you're able to deal with based on the file size.

 

[MrRandizzle]

Any time you record using a quality-based target, it's going to depend on the complexity of the source video. Some content is easier to encode, some content takes more. This will have an effect on how big the overall filesize is given a set CQP value, but generally the more important factor is in the actual quality level you're looking for.

The recommendation of 14-16 for CQP is usually for editing-level quality, but depending on what you're actually using the recording for, this may not be wanted. Use your own eyes and do your own testing -- try higher CQP values, check the quality yourself, and adjust accordingly until you have the quality level that you actually want, or at least something you're able to deal with based on the file size.

I'm looking to upload 1080p60 Let's Plays on YouTube (so original, right?!) And I've been hoping to record game + game audio and voice + face as two seperate files using two instances of OBS. I've also heard about H.265, should I try that, or could a Ryzen 5 2600 + GTX 1070 TI combo be too weak to do that for two recordings AND a game?

 

[carlmmii]

Something to keep in mind is that youtube will always transcode your video file. So if you're not doing any editing of the recording (at least, to the point of needing high quality), then your target should just be something that is equal to or higher quality than what youtube is transcoding to.

I wouldn't recommend h.265 (this is an absolute pain to set up, and not officially supported by OBS -- it requires a lot of ffmpeg tinkering). Standard Nvenc is more than enough for your needs, and is the least impact on your system -- it also lets you do up to 3 encoding sessions, so you're set there for multiple instances.

 

[MrRandizzle]

Something to keep in mind is that youtube will always transcode your video file. So if you're not doing any editing of the recording (at least, to the point of needing high quality), then your target should just be something that is equal to or higher quality than what youtube is transcoding to.

I wouldn't recommend h.265 (this is an absolute pain to set up, and not officially supported by OBS -- it requires a lot of ffmpeg tinkering). Standard Nvenc is more than enough for your needs, and is the least impact on your system -- it also lets you do up to 3 encoding sessions, so you're set there for multiple instances.

Alright, sounds good! So what does increasing the CQ ACTUALLY do? Does it just scale down bitrate, or is it something more complicated than that?

 

[carlmmii]

Quite a bit more complicated. Big-picture-wise though, it targets a specific percentage of perceived quality compared to the original. The lower the CQP number, the lower the difference in quality, which in turn means a higher output quality.

The encoder will use as much bitrate that it needs to achieve this target quality level. So, for lower CQP values, it takes a larger amount of data to reach that quality.

It doesn't have a set bitrate scale it follows when it does this either -- depending on the complexity of the scene, it could be very easy to achieve the required quality level with very little data (i.e. just a black screen). In other cases, hitting the target quality could take an insane amount of data (i.e. crisp fully textured high-motion foliage).

 

[koala]

CQP mode works in a way so that from every video frame some detail is removed to get less information to encode, so the resulting file is smaller. The amount of detail to remove is given with the CQ parameter. The higher it is, the more detail is removed. 0 is no detail removed (=lossless), 50 (the maximum) is "all detail removed".
Bitrate is not relevant for this. The encoder uses as much bitrate as it is required to fulfill the CQ requirement.

Between 0-15, the human brain is unable to tell any difference. From 16 to 20 you start to see a change between the original and the encoded video, although you cannot really tell what is the original or the encoded video. From 21-24 you can tell a video is compressed, and from 25 onwards you clearly see encoding artifacts. Which is not bad by itself - it depends on what you want to do with the video.

A difference of 3 in the CQ parameter means about half (or double) the resulting file size.

CQP mode is actually the generic encoder mode. Internally, every h.264 encoder uses CQP, even for bitrate-orientated output modes. It just varies the CQ parameter internally for every frame. If the bitrate is too high, CQ is tuned down automatically until the averaged bitrate over 10-20 frames is within required bitrate. If the material gets compressed better, so the bitrate isn't fully utilized, CQ is tuned up again.

 

[MrRandizzle]

CQP mode works in a way so that from every video frame some detail is removed to get less information to encode, so the resulting file is smaller. The amount of detail to remove is given with the CQ parameter. The higher it is, the more detail is removed. 0 is no detail removed (=lossless), 50 (the maximum) is "all detail removed".
Bitrate is not relevant for this. The encoder uses as much bitrate as it is required to fulfill the CQ requirement.

Between 0-15, the human brain is unable to tell any difference. From 16 to 20 you start to see a change between the original and the encoded video, although you cannot really tell what is the original or the encoded video. From 21-24 you can tell a video is compressed, and from 25 onwards you clearly see encoding artifacts. Which is not bad by itself - it depends on what you want to do with the video.

A difference of 3 in the CQ parameter means about half (or double) the resulting file size.

CQP mode is actually the generic encoder mode. Internally, every h.264 encoder uses CQP, even for bitrate-orientated output modes. It just varies the CQ parameter internally for every frame. If the bitrate is too high, CQ is tuned down automatically until the averaged bitrate over 10-20 frames is within required bitrate. If the material gets compressed better, so the bitrate isn't fully utilized, CQ is tuned up again.

So I take it that a 1-minute long video with Video Input Capture (iPhone XR through an app called iVCam) and Mic Capture (Samson G-Track Pro and audio bitrate set to 320) is not unusual to come in at 1.18GB? I tried this again after I posted the original post with all the same settings, but changed the scene to include Game Capture (Minecraft) and Display Audio and it was also 1.18GB, but looked kinda choppy when I played it back. Would VBR be viable at all?

 

[koala]

The file size depends on the content of the video, on the resolution, on the frame rate, and on the CQ parameter. Just increase the CQ value and see how the video will look like. Iincrease it until you're fine with the size while keeping the quality still high enough.
You also set Settings->Advanced->Video->Color Range from Partial to Full, this increases file size by about 15%.

Don't use a different rate control. Instead, adjust the CQ parameter until you're satisfied with the file size. Every other rate control will result in worse quality, especially with fast motion parts.

 

[MrRandizzle]

Update: I lowered the CQ to 20 and turned the color range down to partial and I still feel pretty satisfied with the quality, especially considering it will have to go through the YouTube wringer anyways. I also realized that the reason my face-cam + mic audio recordings are so large is because of the Color Correction and Sharpen filters I put on the webcam that seem to double the size. But I've decided that it's worth it and I'll just deal with it. The game capture files are a totally manageable size, so that's a win in my book! Thanks so much for your guys' help! One step closer!

 

[agenthunk]

So 18 cqp is the sweet spot for 1080p 60fps gameplay?

 

[konsolenritter]

There is no common sweep-spot. Read the second post in this thread again. The trade-off between quality and filesize depends on your flavour...

 

[agenthunk]

so h265 is mainly for AMD and the h 264s encoders are usually for nvidia????