Question / Help Smearing and pixelation on recordings with, supposedly, "best settings"

[G_Creature]

Log: https://hastebin.com/qujumowiho

Smearing in question:

Video, just in case: https://streamable.com/r48mo

Screencap from Sony Vegas:

 

[___Maggi___]

Hi,

try CRF instead of CBR. I personally love CRF for local recording! Just switch to CRF, don't change the default value of 23 and set the preset to ultrafast. Give it a try!!!

 

[Zidakuh]

CFR 23 is okay, but not "as good as it gets". My recommendation will be CRF 18 for 720p, CRF 20 for 1080p.
Going over 16 will practically be considered lossless, but takes up huge amounts of space.
Also, a thing to note: video encoded with x264/NVENC/VCE-AMF will allways have a slightly worse quality than the original. This cannot be changed, as that is just how it works.

 

[G_Creature]

I don't have CRF. I have CBR, VBR, CQP and lossless.

 

[BK-Morpheus]

Ah NVENC user...see the comment above from Zidakuh.
The NVENC pendant to x264 CRF should be CQP then. Try CQP 16 maybe

 

[Zidakuh]

CQP is "constant quantizer", meaning that every frame will be compressed the same, regardless of bitrate and complexity.
CRF is "constant ratefactor", meaning the compressor tries to compress every frame to a "quality based of human visual" depending on how complex the image is.
How they both works are very different, but they will achieve the same goal, given enough "quality" (e.g. 16 or 19, lower being better)
While they are fundamentally different, they are both good for HQ recordings.
However, CRF tends to give the best percieved quality.

 

[koala]

However, CRF tends to give the best percieved quality.

Unfortunately, this statement is misleading. Correct would be: "If you set the CQP parameter and the CRF parameter to values that result in percieved similar quality, CRF will produce smaller file sizes".

CRF is made to optimize compression over constant quantizer methods while not sacrificing quality as perceived by the human eye.

Since every current encoder in quality based rate control modes is able to produce quality indistinguishable from the original, it's all about file size, not quality.

 

[Zidakuh]

Unfortunately, this statement is misleading. Correct would be: "If you set the CQP parameter and the CRF parameter to values that result in percieved similar quality, CRF will produce smaller file sizes".

CRF is made to optimize compression over constant quantizer methods while not sacrificing quality as perceived by the human eye.

Since every current encoder in quality based rate control modes is able to produce quality indistinguishable from the original, it's all about file size, not quality.

My bad. I will note this for future reference.

 

[koala]

While we are at it, all those rate control modes (CBR, VBR, CRF, CQP, ICQ etc.) are not fundamentally different from each other. They all have in common that they use exactly the same algorithm to compress a video frame. And they all have a parameter that tells the algorithm: "remove some amount of detail from the frame". The more detail removed, the smaller the compression result.

This parameter, called quantizer, is what is controlled differently across the rate control modes.

• In CQP, the user provides a plain value. Thus, from each frame the same amount of detail is removed.
• In CRF and ICQ, the user provides a guiding value. Additionally, an algorithm in the encoder analyses the frames over time and detect how much motion is present currently. In high motion scenes, it dynamically raises the quantizer to remove a small amount of detail, because the human eye wouldn't notice it. And in low motion scenes, it dynamically lowers the quantizer to add some detail, because the human eye is able to detect more details in low motion scenes.
• In CBR, the user provides a bitrate and no quantizer. An algorithm within the encoder varies the internally used quantizer until the compression just produces frames which, if transmitted over a connection, occupy the provided bitrate. If the algorithm detects the compressed frames are too big, and the bitrate required to transmit is higher than the desired bitrate, it raises the quantizer in future frames to remove more detail and get smaller compressed frames. And if the algorithm detects that there is free bandwidth, it lowers the quantizer in future frames, so the compressed frames get bigger.
• VBR is essentially like CBR, only the bitrate is allowed to vary between a user-provided lower and upper bound, not exactly vary around a single bitrate.