YouTube destroys game footage - please help with bad quality after upload

[tomasz86]

Hello everyone. This is my first posting here, and I am a newbie when it comes to video editing, so please forgive me for any mistakes.

I am trying to record and upload to YouTube a video game footage. The game is Diablo II which runs full screen in 800x600. I am using OBS Studio to record it almost losslessly in x264 (using ffmpeg with qp=0, but the color format is NV12).

The exact specs of the video are as follows.

General
Complete name                  : Diablo II (800x600).mkv
Format                         : Matroska
Format version                 : Version 4 / Version 2
File size                      : 133 MiB
Duration                       : 30 s 48 ms
Overall bit rate mode          : Variable
Overall bit rate               : 37.1 Mb/s
Writing application            : Lavf58.17.101
Writing library                : Lavf58.17.101
ErrorDetectionType             : Per level 1

Video
ID                             : 1
Format                         : AVC
Format/Info                    : Advanced Video Codec
Format profile                 : High 4:4:4 Predictive@L3.1
Format settings                : CABAC / 2 Ref Frames
Format settings, CABAC         : Yes
Format settings, ReFrames      : 2 frames
Format settings, GOP           : M=1, N=15
Codec ID                       : V_MPEG4/ISO/AVC
Duration                       : 30 s 33 ms
Width                          : 800 pixels
Height                         : 600 pixels
Display aspect ratio           : 4:3
Frame rate mode                : Constant
Frame rate                     : 30.000 FPS
Color space                    : YUV
Chroma subsampling             : 4:2:0
Bit depth                      : 8 bits
Scan type                      : Progressive
Writing library                : x264 core 155 r2901 7d0ff22
Encoding settings              : cabac=1 / ref=2 / deblock=1:0:0 / analyse=0x1:0x131 / me=umh / subme=9 / psy=0 / mixed_ref=1 / me_range=24 / chroma_me=1 / trellis=0 / 8x8dct=1 / cqm=0 / deadzone=21,11 / fast_pskip=0 / chroma_qp_offset=0 / threads=6 / lookahead_threads=1 / sliced_threads=0 / nr=0 / decimate=1 / interlaced=0 / bluray_compat=0 / constrained_intra=0 / bframes=0 / weightp=2 / keyint=15 / keyint_min=8 / scenecut=40 / intra_refresh=0 / rc=cqp / mbtree=0 / qp=0
Default                        : No
Forced                         : No
Color range                    : Full
Matrix coefficients            : BT.709

Audio
ID                             : 2
Format                         : FLAC
Format/Info                    : Free Lossless Audio Codec
Codec ID                       : A_FLAC
Duration                       : 30 s 48 ms
Bit rate mode                  : Variable
Channel(s)                     : 2 channels
Channel positions              : Front: L R
Sampling rate                  : 48.0 kHz
Bit depth                      : 16 bits
Default                        : Yes
Forced                         : No

The first problem is that YouTube does not support the 800x600 resolution and downscales the video to 640x480, which is very blurry and looks horrible overall. To overcome the problem I have been trying to upscale the original video to 1600x1200 before uploading. I have tried to use all the different resize algorithms in AviSynth (http://avisynth.nl/index.php/Resize), but it is all in vain, because YouTube still completely butchers the video. In the last few days I have learnt the basics of AviSynth and have been struggling with different settings to preprocess the video specifically for YouTube, but I have been unable to achieve any meaningful results.

Here are a few screen shots of the different videos for more direct comparison.

1) 800x600 (original lossless)

https://drive.google.com/open?id=1m2LiqwQAh3-pZDSm-eqmHcShEy8c3H5w (use the download arrow button, not the YouTube player)


[Attachment 46230 - Click to enlarge]


2) 640x480 (downscaled by YouTube)

https://youtu.be/pzNxizVSLGk


[Attachment 46231 - Click to enlarge]


3) 1600x1200 (upscaled with BlackmanResize)


[Attachment 46232 - Click to enlarge]


4) 1440x1080 (upscaled with BlackmanResize to 1600x1200, downscaled by YouTube)

https://youtu.be/DXX9o7clPnw


[Attachment 46233 - Click to enlarge]


I would be extremely grateful if someone more knowledgeable could play with the original video or simply suggest proper configuration to use with AviSynth (or other programs) so that I can test it myself. I am aware of the fact that YouTube uses very low (or too low) bitrate to maintain proper quality for each resolution, but at the moment the difference is just too much .

[KarMa]

Firstly it looks like you have elevated blacks in your youtube snapshots (black is not black), which implies you might be capturing PC levels instead of TV levels which can be changed in OBS. I would also try Limited/Partial Color range in OBS. For the sake of compression, maybe look into having a less detailed ground as that moving grass eats up a lot of bandwidth and serves little purpose.

Another problem you will face is if you are capturing your content with Rec.601 (Standard for digital SD) but then upscaling that to 720p or higher, then your content will be decoded as Rec.709 even though it's actually still Rec.601. Which can result in some of your problems with muddy colors. I don't know the correct answer but maybe capture your SD content to Rec.709 as long as you plan on upscaling it.

[pandy]

Use pixel art resizer ( https://en.wikipedia.org/wiki/Pixel-art_scaling_algorithms ) to keep/preserve original look - ffmpeg offer for example decent xbr resizer, secondly you can uspcale video to 4k with letterbox - this should allocate more bitrate for video.

You must upscale video anyway to preserve full chroma bandwidth (YT will use YCbCr 4:2:0 and this will hurt video quality).
It is quite late now so i can try to create ffmpeg script tomorrow.

[tomasz86]

First of all, thank you for all your suggestions.

Firstly it looks like you have elevated blacks in your youtube snapshots (black is not black), which implies you might be capturing PC levels instead of TV levels which can be changed in OBS. I would also try Limited/Partial Color range in OBS. For the sake of compression, maybe look into having a less detailed ground as that moving grass eats up a lot of bandwidth and serves little purpose.

Another problem you will face is if you are capturing your content with Rec.601 (Standard for digital SD) but then upscaling that to 720p or higher, then your content will be decoded as Rec.709 even though it's actually still Rec.601. Which can result in some of your problems with muddy colors. I don't know the correct answer but maybe capture your SD content to Rec.709 as long as you plan on upscaling it.

Hmm, I did set OBS to record with Rec.709 but with full colour range. I have now changed it to 709/limited and prepared a new recording. The game does not have many graphics related settings though.

Before:


[Attachment 46239 - Click to enlarge]


After:


[Attachment 46240 - Click to enlarge]


In the new recording I have changed the Lightning Quality from "high" to "low". The most visible difference is that the grass looks flatter and less detailed, and there is no rain any more. I have uploaded it to YouTube to check if there is any difference. I am unsure about the results though.

Do in-game gamma / contrast settings play any role when it comes to YouTube compression?

1) 800x600 (original lossless)

https://drive.google.com/open?id=1AkIL9SyrSJCcdvAuHlSKaZ3dP8AayEJL (use the download arrow button, not the YouTube player)


[Attachment 46241 - Click to enlarge]


2) 640x480 (800x600 downscaled by YouTube)

https://youtu.be/Ui2w4EG80E4


[Attachment 46242 - Click to enlarge]


3) 1600x1200 (800x600 upscaled with BlackmanResize)


[Attachment 46244 - Click to enlarge]


4) 1440x1080 (800x600 upscaled to 1600x1200 with BlackmanResize, downscaled by YouTube)

https://youtu.be/o8KVy4K6gaU


[Attachment 46243 - Click to enlarge]


The new recording does seem to look a little bit better in 1440x1080 on YouTube, although there is still a big difference when compared to the original. The 640x480 version always looks horrible, but have already given up on the low resolution ones. I would just be happy to have the 1080p version look decent.

Is there anything else that I could do to tweak the video for YouTube?

Use pixel art resizer ( https://en.wikipedia.org/wiki/Pixel-art_scaling_algorithms ) to keep/preserve original look - ffmpeg offer for example decent xbr resizer, secondly you can uspcale video to 4k with letterbox - this should allocate more bitrate for video.

You must upscale video anyway to preserve full chroma bandwidth (YT will use YCbCr 4:2:0 and this will hurt video quality).
It is quite late now so i can try to create ffmpeg script tomorrow.

I tried to use NNEDI3 before, and have now tried the -filter:v "xbr=2" with ffmpeg, but the resulting image is very "plasticky" (which was also the case with NNEDI3).

5) 1600x1200 (800x600 upscaled with ffmpeg xbr)


[Attachment 46245 - Click to enlarge]


6) 1440x1080 (800x600 upscaled with ffmpeg xbr, downscaled by YouTube)

https://youtu.be/zfVg7RSlLbE


[Attachment 46246 - Click to enlarge]


At the moment, I personally like the blackman upscaled 4) version better though. As far as I understand letterboxing, it would create permanent black bars, right? I would like to avoid doing that, if possible.

[manono]

Can't you just upsize to a YouTube approved resolution so they don't resize everything you send them? Those might include 960x720 and 1440x1080. Why send them 1600x1200 when you know they're just going to resize it yet again?

[KarMa]

Most youtubers under 10k subs are usually just stuck with H.264 encodings of their videos which gives worse quality for the bitrate. Uploading videos in 4K seems to force youtube to encode videos with VP9, giving better quality encodings.

[pandy]

I tried to use NNEDI3 before, and have now tried the -filter:v "xbr=2" with ffmpeg, but the resulting image is very "plasticky" (which was also the case with NNEDI3).

Well - this is particular type of game so must decide or you trying to make look it more modern or look closer to original. I know that pixel art resizers are preferred by retro community so that's why i suggested to use it.

At the moment, I personally like the blackman upscaled 4) version better though. As far as I understand letterboxing, it would create permanent black bars, right? I would like to avoid doing that, if possible.

Ok i think first you must accept that you can't control YT encoding settings - they are under YT control, they use own H.264 encoder (probably heavily modified x264) and own VP9 encoder - all you can control is resolution - all lower resolutions are created by YT and you loosing ability to control anything there (they may even use plain bilinear resizer).

Secondly your source is very difficult to encode - almost every CG source is prone to aliasing - it is very sharp, unnatural and as such almost impossible to encode, side to this usually heavy dithering is used (limited color palette) and dithering from signal processing perspective is noise - noise can't be compressed due high entropy. Additionally your source is RGB 4:4:4 (converted to YCbCr 4:4:4) - and YT deliver only 4:2:0 format where chroma bandwidth is reduced by half.
All this makes game very difficult content for high quality YT video.

Blackman has lowest rigging amongst all kernel sinc based filters however it is still worse than spline family filters - strongly advise to use spline resizers.
Letterbox doesn't must be always black bars (pillar box in case of 4:3 on 16:9) - you can always use overlay with aspect ratio distorted background with heavily blurred source and placed on top normal (correct aspect) video - such videos are common way to deal with pillar box - real life closest thing to such approach is Philips ambilight seen commonly in Philips TV's.

If i will be on your place then i would:

1. resize video with xbr (edi) class filter 4 times
2. upscale video with spline to 2880x2160
3. create layer with video upscaled to 3840x2160
4. apply strong blur (horizontal direction) to layer 3840x2160
5. overlay video 2880x2160 on top of background video (3840x2160)
6. encode such video with fast or very fast x264 preset and sufficiently high bitrate (your video will be recompressed by YT anyway)

Above this i would consider to perform colour correction if necessary (colour space conversion and level quantization - they must be determined by you and set explicitly to avoid faulty automatics).
You may wish to add small temporal blur that may reduce impact of YT reencoding.
Consider to use other services than YT - for example Vimeo offers possibility to download original source. Usually those services offers marginally higher video quality too.

[tomasz86]

Can't you just upsize to a YouTube approved resolution so they don't resize everything you send them? Those might include 960x720 and 1440x1080. Why send them 1600x1200 when you know they're just going to resize it yet again?

This is actually a very good question. I have read in many places that upscaling a video to a resolution higher than 1152 px vertically tricks the YouTube algorithm to use higher bitrates to encode it. After all my testing, however, I am not sure there is any benefit though.

I did test direct upscaling to both 720p and 1080p, but there was barely any differences to upscaling to 1600x1200 and then letting YouTube downscale the videos to 1080p (and lower). I have just done a new test though, and the results are below.

1) 1440x1080 (800x600 upscaled with BlackmanResize)


[Attachment 46249 - Click to enlarge]


2) 1440x1080 (800x600 upscaled with BlackmanResize, uploaded to YouTube)


[Attachment 46250 - Click to enlarge]

https://youtu.be/rsysNHKHgbM


3) 1440x1080 (800x600 upscaled to 1600x1200 with BlackmanResize, downscaled by YouTube) (same video as 4) in my previous post)

https://youtu.be/o8KVy4K6gaU


[Attachment 46243 - Click to enlarge]

Most youtubers under 10k subs are usually just stuck with H.264 encodings of their videos which gives worse quality for the bitrate. Uploading videos in 4K seems to force youtube to encode videos with VP9, giving better quality encodings.

During my testing, I did force YouTube to offer a VP9 version by upscaling the video to 2400x1800. The video was listed as "1440p" and actually looked decent in that resolution. The problem is that the downscaled 1080p version looked even more blurry than when downscaling from 1600x1200. I also doubt that many people even watch YouTube videos in those higher resolutions, because a) their monitor resolution is still much lower, and b) the VP9 codec itself is much more CPU heavy (or requires GPU hardware acceleration) than the old x264, and a typical home PC lacks power to play those VP9 encoded videos anyway.

That is why I am just trying to improve the video quality in the standard x264 1080p version (with its low bitrate).

Secondly your source is very difficult to encode - almost every CG source is prone to aliasing - it is very sharp, unnatural and as such almost impossible to encode, side to this usually heavy dithering is used (limited color palette) and dithering from signal processing perspective is noise - noise can't be compressed due high entropy. Additionally your source is RGB 4:4:4 (converted to YCbCr 4:4:4) - and YT deliver only 4:2:0 format where chroma bandwidth is reduced by half.
All this makes game very difficult content for high quality YT video.

Thank you for a detailed explanation. I will try to experiment with the colour settings. I guess that the fewer colours, the less bandwith the video will use. Is this correct?

Blackman has lowest rigging amongst all kernel sinc based filters however it is still worse than spline family filters - strongly advise to use spline resizers.

I chose Blackman after testing all the different resizing filters as it had definitely less distortion than plain Sinc or Lanczos. The effect when using spline filters was similar, so I will check them again, but the YouTube quality was also basically the same, unfortunately.

Letterbox doesn't must be always black bars (pillar box in case of 4:3 on 16:9) - you can always use overlay with aspect ratio distorted background with heavily blurred source and placed on top normal (correct aspect) video - such videos are common way to deal with pillar box - real life closest thing to such approach is Philips ambilight seen commonly in Philips TV's.

If i will be on your place then i would:

1. resize video with xbr (edi) class filter 4 times
2. upscale video with spline to 2880x2160
3. create layer with video upscaled to 3840x2160
4. apply strong blur (horizontal direction) to layer 3840x2160
5. overlay video 2880x2160 on top of background video (3840x2160)
6. encode such video with fast or very fast x264 preset and sufficiently high bitrate (your video will be recompressed by YT anyway)

Above this i would consider to perform colour correction if necessary (colour space conversion and level quantization - they must be determined by you and set explicitly to avoid faulty automatics).
You may wish to add small temporal blur that may reduce impact of YT reencoding.

I will try to experiment with these settings, although I am not sure whether I will manage to do all of this with my still very limited knowledge. Thank you anyway for all the suggestions. I will try to do what I can and report later.

Consider to use other services than YT - for example Vimeo offers possibility to download original source. Usually those services offers marginally higher video quality too.

That is probably true, but YouTube has basically monopoly on online video. I have personally never seen anyone actually watching videos on those other websites.

[pandy]

Thank you for a detailed explanation. I will try to experiment with the colour settings. I guess that the fewer colours, the less bandwith the video will use. Is this correct?

Oposite - less colours equal more dithering and more CG like look i.e. lower encoder efficiency - commonly used encoders prefer smooth, continuous gradient content without sudden (high contrast) changes. You may try to apply on enlarged video filters like gradfun with very strong settings - they will change video but this change will make video easier to compress.

I chose Blackman after testing all the different resizing filters as it had definitely less distortion than plain Sinc or Lanczos. The effect when using spline filters was similar, so I will check them again, but the YouTube quality was also basically the same, unfortunately.

Theoretically (mathematically) perfect is pointresize video (however factor must be integer 2, 3, 4 etc) - sometimes i use pointresize, apply large kernel median filter and downscale video to reduce perceived "pixelity".
Sinc kernels are always "ringing" as sinc function is "ringing" - also limiting window ( https://en.wikipedia.org/wiki/Window_function ) may be "ringing" type

I will try to experiment with these settings, although I am not sure whether I will manage to do all of this with my still very limited knowledge. Thank you anyway for all the suggestions. I will try to do what I can and report later.

It should be not difficult - i see you have skills so this require only some time - good luck

That is probably true, but YouTube has basically monopoly on online video. I have personally never seen anyone actually watching videos on those other websites.

We decide what is monopoly - you can always provide link to both (YT and Vimeo).

--
testes crude temporal blur (ffmpeg) and it doesn't look very bad...

tmix=frames=6:weights='.004 .012 .037 .111 .333 1'

[tomasz86]

OK, so before trying any more advanced editing, I have decided to take a few steps back.

First of all, Diablo II in single player is hard locked to 25 fps. However, recording in 25 fps in OBS Studio with a 60 Hz monitor results in stuttering, so I was forced to record in 30 fps. I have now managed to fix this problem by setting my monitor to 50 Hz (which is 25 multiplied by 2). This way I am finally able to record the game in its native 25 fps which will also hopefully give a slight boost in quality on YouTube (since there will be fewer frames overall for the given bitrate).

Secondly, I re-tested most of the resizing algorithms used in AviSynth. Also, I wanted to see whether anti aliasing made any difference for the YouTube's compression. I found the Santiag anti aliasing script (http://avisynth.nl/index.php/Santiag) which I applied to the videos. The script allows for defining different anti aliasing algorithms, so I tested both NNEDI3 and EEDI3.

Lastly, I uploaded all videos to YouTube, downloaded the 1080p version of each of them, and then compared with the originals. At the moment, the comparison is purely subjective (i.e. done with my eyes ), but I am now looking into tools to automatically compare and calculate differences between images. I have found ImageMagick (http://www.imagemagick.org), but please let me know if there is other software more suitable for this.

I have compared both full screen captures, and also cropped images (focused on the characters and monsters, with not much terrain and trees). I personally like the anti aliased versions in each case more, and among them specifically the EEDI3 ones. Anti aliasing also does seem to help for better quality in the YouTube compression. Strangely enough, I think that the images resized with Sinc and anti aliased with EEDI3 eventually came up the best looking on YouTube...


(WARNING: a lot of images!)


1) Bilinear

Original

[Attachment 46262 - Click to enlarge]

[Attachment 46286 - Click to enlarge]

YouTube

[Attachment 46263 - Click to enlarge]

[Attachment 46287 - Click to enlarge]

Original AA NNEDI3

[Attachment 46264 - Click to enlarge]

[Attachment 46288 - Click to enlarge]

YouTube AA NNEDI3

[Attachment 46265 - Click to enlarge]

[Attachment 46289 - Click to enlarge]

Original AA EEDI3

[Attachment 46266 - Click to enlarge]

[Attachment 46290 - Click to enlarge]

YouTube AA EEDI3

[Attachment 46267 - Click to enlarge]

[Attachment 46291 - Click to enlarge]


2) Bicubic

Original

[Attachment 46268 - Click to enlarge]

[Attachment 46292 - Click to enlarge]

YouTube

[Attachment 46269 - Click to enlarge]

[Attachment 46293 - Click to enlarge]

Original AA NNEDI3

[Attachment 46270 - Click to enlarge]

[Attachment 46294 - Click to enlarge]

YouTube AA NNEDI3

[Attachment 46271 - Click to enlarge]

[Attachment 46295 - Click to enlarge]

Original AA EEDI3

[Attachment 46272 - Click to enlarge]

[Attachment 46296 - Click to enlarge]

YouTube AA EEDI3

[Attachment 46273 - Click to enlarge]

[Attachment 46297 - Click to enlarge]


3) Blackman

Original

[Attachment 46274 - Click to enlarge]

[Attachment 46276 - Click to enlarge]

YouTube

[Attachment 46275 - Click to enlarge]

[Attachment 46277 - Click to enlarge]

Original AA NNEDI3

[Attachment 46278 - Click to enlarge]

[Attachment 46280 - Click to enlarge]

YouTube AA NNEDI3

[Attachment 46279 - Click to enlarge]

[Attachment 46281 - Click to enlarge]

Original AA EEDI3

[Attachment 46282 - Click to enlarge]

[Attachment 46284 - Click to enlarge]

YouTube AA EEDI3

[Attachment 46283 - Click to enlarge]

[Attachment 46285 - Click to enlarge]


4) Gauss

Original

[Attachment 46298 - Click to enlarge]

[Attachment 46300 - Click to enlarge]

YouTube

[Attachment 46299 - Click to enlarge]

[Attachment 46301 - Click to enlarge]

Original AA NNEDI3

[Attachment 46302 - Click to enlarge]

[Attachment 46304 - Click to enlarge]

YouTube AA NNEDI3

[Attachment 46303 - Click to enlarge]

[Attachment 46306 - Click to enlarge]

Original AA EEDI3

[Attachment 46307 - Click to enlarge]

[Attachment 46309 - Click to enlarge]

YouTube AA EEDI3

[Attachment 46308 - Click to enlarge]

[Attachment 46310 - Click to enlarge]


5) Point

Original

[Attachment 46311 - Click to enlarge]

[Attachment 46313 - Click to enlarge]

YouTube

[Attachment 46312 - Click to enlarge]

[Attachment 46314 - Click to enlarge]

Original AA NNEDI3

[Attachment 46315 - Click to enlarge]

[Attachment 46317 - Click to enlarge]

YouTube AA NNEDI3

[Attachment 46316 - Click to enlarge]

[Attachment 46318 - Click to enlarge]

Original AA EEDI3

[Attachment 46319 - Click to enlarge]

[Attachment 46321 - Click to enlarge]

YouTube AA EEDI3

[Attachment 46320 - Click to enlarge]

[Attachment 46322 - Click to enlarge]


6) Spline16

Original

[Attachment 46323 - Click to enlarge]

[Attachment 46325 - Click to enlarge]

YouTube

[Attachment 46324 - Click to enlarge]

[Attachment 46326 - Click to enlarge]

Original AA NNEDI3

[Attachment 46327 - Click to enlarge]

[Attachment 46329 - Click to enlarge]

YouTube AA NNEDI3

[Attachment 46328 - Click to enlarge]

[Attachment 46330 - Click to enlarge]

Original AA EEDI3

[Attachment 46331 - Click to enlarge]

[Attachment 46333 - Click to enlarge]

YouTube AA EEDI3

[Attachment 46332 - Click to enlarge]

[Attachment 46334 - Click to enlarge]

7) Spline36

Original

[Attachment 46335 - Click to enlarge]

[Attachment 46337 - Click to enlarge]

YouTube

[Attachment 46336 - Click to enlarge]

[Attachment 46338 - Click to enlarge]

Original AA NNEDI3

[Attachment 46339 - Click to enlarge]

[Attachment 46341 - Click to enlarge]

YouTube AA NNEDI3

[Attachment 46340 - Click to enlarge]

[Attachment 46342 - Click to enlarge]

Original AA EEDI3

[Attachment 46343 - Click to enlarge]

[Attachment 46345 - Click to enlarge]

YouTube AA EEDI3

[Attachment 46344 - Click to enlarge]

[Attachment 46346 - Click to enlarge]


8) Spline64

Original

[Attachment 46347 - Click to enlarge]

[Attachment 46349 - Click to enlarge]

YouTube

[Attachment 46348 - Click to enlarge]

[Attachment 46350 - Click to enlarge]

Original AA NNEDI3

[Attachment 46351 - Click to enlarge]

[Attachment 46353 - Click to enlarge]

YouTube AA NNEDI3

[Attachment 46352 - Click to enlarge]

[Attachment 46354 - Click to enlarge]

Original AA EEDI3

[Attachment 46355 - Click to enlarge]

[Attachment 46357 - Click to enlarge]

YouTube AA EEDI3

[Attachment 46356 - Click to enlarge]

[Attachment 46358 - Click to enlarge]


9) Sinc

Original

[Attachment 46359 - Click to enlarge]

[Attachment 46361 - Click to enlarge]

YouTube

[Attachment 46360 - Click to enlarge]

[Attachment 46362 - Click to enlarge]

Original AA NNEDI3

[Attachment 46363 - Click to enlarge]

[Attachment 46365 - Click to enlarge]

YouTube AA NNEDI3

[Attachment 46364 - Click to enlarge]

[Attachment 46366 - Click to enlarge]

Original AA EEDI3

[Attachment 46367 - Click to enlarge]

[Attachment 46369 - Click to enlarge]

YouTube AA EEDI3

[Attachment 46368 - Click to enlarge]

[Attachment 46370 - Click to enlarge]

[jagabo]

Try HQ2X()+Santiag().