Bitrate comparison test

doublevr

We have been doing fundamental bitrate research for SLR streaming encodings. We discovered a golden ratio 30Mbps variable bitrate with “highest quality” setting after which you don't perceive quality difference with any extra bitrate for most on indoors videos. That's why it takes days for SLR to encode such video. It's a slow, but super efficient process. It equals to 60Mbps same time encoding.

There is no such thing as "low" bitrate that is frequently discussed as the algorithms decide how much bitrate is required for each frame to deliver the best visuals which is very different to constant bitrate output from the camera.

We don't set the actual bitrate, but the maximum quality setting, with algorithmic bitrate for each video segment depending on the complexity. If we will lock bitrate to some specific number then one part of the scene will always look good because of the low complexity and another one will be bad because of the high complexity.
This test will not reflect our encoding approach and will be completely artificial.

Right now you can take original quality file and the best resolution encoding and tell me if you see the difference.

mirdumar

The problem is that we don't set encoding bitrates. We set desired quality and encoder automatically picks the best bitrate for each video segment depending on the complexity to achieve constant quality.

Partially true, you use crf=15, which would result in almost lossless file, but you also use maxrate=30000, which limits bitrate to 30mbps.

This is called variable bitrate encoding. If we will lock bitrate to some specific number then one part of the scene will always look good because of the low complexity and another one will be bad because of the high complexity.

But you lock bitrate to some specific number, as i said you use maxrate=30000 (for some studios you even use 22000), which automatically prevents bitrate to go above that limit, and you end up with constant bitrate file, high resolution scenes need more than 30mbps.

g2kbuffetboy

doublevr Right now you can take original quality file and the best resolution encoding and tell me if you see the difference.

Hmm, I believe I already provided comparisons for this a couple years back (October 2020)
https://forum.sexlikereal.com/d/1091-can-slr-offer-a-higher-quality-h265-option-for-5k-6k-currently-it-is-21mbps/8

Is this what you're asking for, or something else? With the screenshots I provided in that thread, especially if you look at their forehead between their two eyebrows, you can see the details/bumps on their skin much more distinctly in the original file vs the compressed file. Amplify that by when you are watching in a headset and everything is "zoomed in" and with motion, it makes a noticeable difference for me on a Pimax 8KX.

Witcher Ex 1
Notice the bumps on Whitney's forehead area, her eyes, and nose, etc.
(Original)

(Rendition)

Naughty Dorm Ex 2
This one is really obvious when looking at Melody's face.
(Original)

(Rendition)

More examples in the link thread.

I mean, there is a reason why 4K HDR Bluray discs (128~ Mbps) still exist and are wayyyy better than 4K HDR streaming (Netflix/Disney/HBO at 20~ Mbps) for the Home Theater enthusiasts. The same applies for VR.

spacepirate

mirdumar ive seen people try to explain to him before that slrs vids are in fact constant bitrate despite slr choosing a vbr / constant quality option when encoding but i dont think he gets it. keeps trying to explain video encoding to us like we are toddlers.

someoneX

mirdumar True, the fact that almost every encoding results in a 30 Mbps file, even though you are using variable bitrate encoding, shows that your setting introduces a loss of quality and that your upper limit (30 Mbps) is too low.

If 30 Mbps were sufficient as an upper limit, you should get files with different bit rates depending on the scene.

If you want almost lossless encodings, you would have to raise the "maxrate" value until you regularly get files that don't reach that upper limit.

Your current settings result in a pseudo-fixed-bitrate file, because you hit the upper limit 99% of the time.

doublevr

Missed there was a maximum cap for the highest resolution.

There's 30Mbps maximum average over 10 seconds time period which roughly corresponds to the player playback cache. This way we can avoid buffering even for very complex parts of the file.

Checking with dev team what are our options.

dango789

Missed there was a maximum cap for the highest resolution.

There's 30Mbps maximum average over 10 seconds time period which roughly corresponds to the player playback cache. This way we can avoid buffering even for very complex parts of the file.

Checking with dev team what are our options.

Okay, the plan is to test Original file to SLR highest encoding with as many people and see if there's a difference. Please let me know how it worked for you.

I LOVE this response. Thank you @doublevr for listening to us and for working with the devs to brainstorm and test solutions.

This thread is very promising.

doublevr

Okay, the plan is to test Original file to SLR highest encoding with as many people and see if there's a difference. Please let me know how it worked for you.

LordCrash

doublevr
Please make sure to use different videos for comparions between different bitrates, like 30,000 kbit/s and 60,000 kbit/s.

Like I said in the other thread, higher bitrate is especially valuable for scene with a lot of movement and "background noise", like e.g. plants and trees etc. in outdoor scenes. The difference isn't that glaring in indoor scenes with a fixed background and generally a lot less movement.

I'd suggest an indoor scene with only one girl, an outdoor scene with foliage in the background and something like a harem scene with a lot of movement by different girls etc. While the difference in the first one might only be noticable in some instances it should be a lot more visible in the latter two, given the overall video quality is already decent enough.

With a flexible, uncapped bitrate scenes with a lot of movement should result in higher bitrates anyway.

fenderwq

doublevr for me the 6k movies usually don't make a huge difference (unless they are particularly well shot).
If you really want to test it I would suggest this movie as it has very good image quality: https://www.sexlikereal.com/scenes/ass-and-you-shall-receive-28336.
If it has to be a SLR original I would suggest: https://www.sexlikereal.com/scenes/big-black-tits-28255.
These clips both have very good image quality which youll especially notice on things like face closeups and off course they are 4000p +.

doublevr

g2kbuffetboy right. Mail these to me at alex@sexlikereal.com or get them uncompressed. It's either Imgur is overcompressing or I see no difference on my screen.

g2kbuffetboy

doublevr They are uncompressed already if the image is opened in a new browser tab. Actually I would recommend to compare within separate tabs too, because you quickly switch between each while staring at the same spot. But I'll email them over to you as well 👍️

LordCrash

doublevr
I actually don't think that screenshots are the best way to measure the results of different bitrates. The bitrate dictates how smooth a video plays and that can hardly be captured in stills. It might be a good starting point but it's not very surprising that the difference in screenshots isn't that glaring.

doublevr

g2kbuffetboy Curious if it makes sense to see these in VR

g2kbuffetboy

doublevr Curious if it makes sense to see these in VR

I think mileage will vary depending on the headset used, and the viewing conditions as well as the lighting conditions of each video. Generally speaking, for my casual viewing sessions I will watch the 6k compressed file more often than the Original file for the sake of ease of use, and because sometimes I like to skip around a lot and I don't want to kill my computer.

But for those times, when I really want to soak in a scene, especially when it's my first time watching it, I prefer the Original file. I noticed a good amount of difference especially with the bigger group scenes, because there is so much more action and movement in each frame. That's when every bit matters, IMO.

bobbytables

If you're going to do this test do it with a colorful scene with gradients. Bitrate issues become a lot more apparent when you start dropping and muddling color when compensating for artefacts. I'd also try it on a greenscreen video, one of the biggest issues people (especially you based on recent scenes) have is accurately eliminating green and a chunk of that is likely due to bitrate.

LordCrash

Since we're actually discussing the issues with open end now I want to re-post something I've already posted in another thread on the topic:

What is Video Bitrate, and how does it Affect Video Quality?
In simple terms, video bitrate refers to the amount of video bits/data transferred within a second. It is vital to note that video bits are just strings of data that make up the video you watch. They’re more like the digital building blocks of your videos.

Since bitrate has to do with the amount of ‘data bits’ being transferred along with the ‘rate’ (i.e., speed) at which the transfer is occurring, it is measured in bits per second (bps). Higher variations include kilobits per second (Kbps) and megabits per second (Mbps).

Basically, when you have a higher bitrate, it means that a larger amount of ‘video bits’ are being uploaded within a second. This naturally leads to a higher quality video relative to its resolution. Also, increasing the bitrate will increase the size of your output video file and result in relatively slower export.

Conversely, lower bitrate results in lower video quality, smaller size, and faster export.

Another thing to understand is that you need to use just the right amount of bitrates, as extremely high bitrates would be wasting your bandwidth, and could potentially affect your hardware.

Bitrate vs. Resolution: Which is More Important for Video Quality?
Video resolution refers to the number of picture elements (pixels) spread across a display unit. A higher concentration of pixels (a.k.a higher resolution) leads to higher visual quality.

I’m sure you’ve heard much more about video resolution than video bitrate. But does that make it more important for video quality?

Well, not exactly.

Basically, higher resolution yields a clearer picture. On the other hand, higher bitrate yields smoother playback. Thus, a high-res video with low bitrate will produce a cringy output. You could say the same for high bitrate and low resolution.

In essence, there has to be a balance between high resolution and high bitrate.

Of course that's only a generalization and VR scenes with little movement shot in an indoor setting with little to no movement in the background probably don't require an ultra-high bitrate every time but there are scenes that benefit substantially by a higher bitrate and a hard cap at 30,000 kbps isn't ideal for best video quality. The more movement and color gradients and stuff you have in a scene the higher the bitrate should be.

sheldoncooper

Can some one provide some gifs

spacepirate

sheldoncooper gifs are even worse

doublevr

Our Unity teamlead is killing it:

we can and we have it in plans (no timeline yet), first we need to do a proper research on what would be the best approach. I would like that caching is supported on all platforms and has the flexibility on how and what we want to cache. Either we find a solution that already provides this or we implement it ourselves, but that will take some time to do it properly.

That's the spirit 🔥

Next Page »