audioMotion Background videos into AV1

[JesusGod-Pope666-Info]

I think I am getting closer to some real settings for this video. Although getting banned from everywhere - but, by testing and looking around and little by little building it up. Here is where I am - now these files do not have audio, but there are some audio settings anyhow in the command - still not certain on those settings on opus 96 or 128 or something entirely else - but here we have a good Quality to Size as far as I can see, the Zip has the AV1 files, you can compare with the originals and see what you think - 17.6MB to 5.3MB:

#!/bin/bash
mkdir -p c36p5g240k30d10o96Ln
for input in *.{mp4,m4a,mkv,webm,avi,flv}; do
[[ -f "$input" ]] || continue
filename="${input%.*}"
output="c36p5g240k30d10o96Ln/${filename}.mp4"
echo "-#-#-#-# ENCODING: $input → $output"
time ffmpeg -hide_banner -loglevel warning -y -i "$input" \
-c:v libaom-av1 -movflags +faststart -pix_fmt yuv420p10le -row-mt 1 \
-crf 36 -b:v 0 -cpu-used 5 -g 240 -keyint_min 30 \
-c:a libopus -b:a 96k       -af "loudnorm=I=-14:LRA=11:TP=-2" \
"$output"
done
echo "-#-#-#-# AV1o c36p5g240k30d10o96Ln Encoding DONE ####"

c36p5g240k30d10LN96.zip

[JesusGod-Pope666-Info]

[JesusGod-Pope666-Info]

I should probably have uploaded both the originals with the encoded so people could easily check and give their opinion.... Kinda would like to hear from people on it.

[JesusGod-Pope666-Info]

Both Originals and Encodings, compare and voice your opinion! :)

Oh no, I made a mistake, BRB with a new file.

[JesusGod-Pope666-Info]

Here we go! :)

Originals and Encodings to compare.zip

[JesusGod-Pope666-Info]

[JesusGod-Pope666-Info]

Weird.... it does not seem the glitch shows up in the SVT encoder version at CFR 36 and 7 in preset....
And AOM should be the better one of the 2, yet.... it seems clean.

Why did it break the image.......

I tried running the AOM at 25 CFR, preset 5, pretty high quality, and the glitch although a lot less, is still showing up - hard to capture.

It is weird because the AOM should be the better one of the two encoders and yet seems to have this weird issue of this glitch. Although the SVT encoder file is somewhat bigger, but still a lot smaller then the AOM 25 CFR, has none of it. Just weird.

[JesusGod-Pope666-Info]

Lets try again, this is the SVT with low settings - and the glitch is not there to be seen.

[JesusGod-Pope666-Info]

Testing SVT a little more - I did try to fix the AOM with experimental 1 but apparently nothing worked. Tried compiling and all, but I do not know, it kinda did not have the features I was trying to get.

Weird enough SVT seems to work more like expected on 1 setting - like more CPU/Preset in AOM, as to say lover number - makes the files bigger, so it does not seem it really compresses more, but also makes the size bigger - kinda weird.
Were as SVT actually do make the files smaller when you use more CPU/Preset - and I guess quality as well.

All these nonsense games of working differently and positive and negative numbers, but apparently very very hard for programmers to have sanity and keep to some solid foundation overall. They just have do reinvent the wheel over and over again.

Higher numbers better quality makes sense, not lowering the number. But this is what we have with ffmpeg AV1 - and apparently CPU/Preset makes 2 different outputs on AOM and SVT, make it make sense.

Now of cause, AOM still have that weird glitch - which SVT just does not have. It is just weird. Trying to make a 0 SVT and see how that works out on Size and Quality, although takes the longest, being the highest setting.

[JesusGod-Pope666-Info]

I did find out that SVT also do make something, although a little difference of glitch at the place - but it is there. More like a black dot thing....

Anyway, it seems I have found out that changing -lag-in-frames might solve the issue - so I am doing some testing on that at the moment, but everything I have made until now, whether smaller then 35 as per default or more seems to solve the glitch. Dunno why..... Weird the default would be the one having the issue. But I am now trying 34 and 36 to see if it is there or disappear.

[JesusGod-Pope666-Info]

Well.... Oddly enough I saw different results on the Glitch but 48 does seem to sort it out - so I would think that it is the most safe setting overall. It does make the file somewhat bigger, but I think we are already so low so it does not matter that much.
Overall with 48 frames seeing ahead, should in theory make it better to calculate stuff and sort things out - and it does remove the glitch which hopefully will make it do less of such in other places. Although I have not really noticed any other place, but if it is one place it could come other places as well. So better sort it out with the 48 setting - although I am still trying things out and it takes a lot of time. Although have put it in action with way to many files to encode. It clearly have its work to do.

I could see how low I could go before the glitch returns - but overall.... if it sorts out other glitches with the extra look ahead of 48, then so be it. It is the max setting in AOM - but it does make the file size bigger.

Anyway..... this has and is a hard one to totally sort out. And takes lots of time making the video and check the quality and size and all.

[JesusGod-Pope666-Info]

Newest settings for the audioMotions default video's - should be a little bigger but less chance of glitches, it should be a little higher quality, but not done any benchmarks on them, comparing them. :

c36p5g240d10LIF48aC.zip

[JesusGod-Pope666-Info]

Overall I am not sure there was any glitches at all in the first batch, but as I found that glitch in a video - and sorted it out with 48 look forward - that is just my default setting to use now. Files get a little bigger but not that much - but should get a slight boost in quality as well. But should be less likely their will be any small glitches with the new settings - so..... That is what I go with.

[JesusGod-Pope666-Info]

darkijah@Darkijah-USBSSD:~/downloads/test/Originals.and.Encodings.to.compare$ find . -maxdepth 1 -type f -printf "%s %f\n"
| numfmt --to=iec --field=1
| sort -k2
681K Flying above clouds.mp4
725K Flying above cloudsNEW.AV1.mp4
3.5M Flying above clouds - ORIGINALS.mp4
1.5M Smoke and Sparks.mp4
1.6M Smoke and SparksNEW.AV1.mp4
5.2M Smoke and Sparks - ORIGINALS.webm
2.1M Starfield.mp4
2.1M StarfieldNEW.AV1.mp4
6.3M Starfield - ORIGINALS.mp4
1.2M Synthwave Landscape.mp4
1.3M Synthwave LandscapeNEW.AV1.mp4
2.8M Synthwave Landscape - ORIGINALS.mp4

[JesusGod-Pope666-Info]

So as can be seen, a little bigger, but not much. And less likely risk of glitches as far as I know. At least I got the glitch gone from the video I found one by upping it to 48 of look ahead, but it does make the files slightly bigger.

[JesusGod-Pope666-Info]

Original, old AV1 CFR 30, new AV1 CFR 36
Sure getting some much better results with these new settings :) At least a lot more in the area of effect that I had hoped for.

Now I did begin with kinda thinking CFR 37 was enough but did raise it to 36 and then the extra FPS foresight fixing the glitch issue - one could go higher or lower from this point but overall, it seems pretty good to me. Like one might squeeze a little down by CFR 37, but..... not sure it is really worth it and I cannot remember the real reason why I went up to 36 - but there was a reason. But we sure are slicing a good amount off by these new settings I am using compared to the old.

Anyway, I think these settings are good overall - here is a picture. Original, CFR 30 and CFR 36:

[JesusGod-Pope666-Info]

Originals 7.1GB - CFR 30 4.8GB - CFR 36 2.4GB
Pretty good reduction in size, and the videos to me looks fine.

[JesusGod-Pope666-Info]

#!/bin/bash 
outdir="c36p5g240d10LIF48cA"; mkdir -p "$outdir"; time {
for input in *.{mp4,m4a,mkv,webm,avi,flv}; do
[[ -f "$input" ]] || continue
filename="${input%.*}"; output="$outdir/${filename}.mp4"
echo "RUN: $input → $output"; ffmpeg -hide_banner -i "$input" \
-c:v libaom-av1 -movflags +faststart -crf 36 -b:v 0 \
-cpu-used 5 -g 240 -pix_fmt yuv420p10le -lag-in-frames 48 \
-c:a copy \
"$output"; done; }; echo "-#-#-#-# AV1 $outdir DONE -#-#-#-# -#-#-#-#"