Hi there.

[JesusGod-Pope666.Info]

Kinda newbie here:

I have this weird glitch showing up in AOM, even at 25 CFR in this movie called Nefarious in the dark shadows. It seems like it copies a table leg maybe and expands it in front of the guy walking in the shadows. I tested with SVT even on very low settings, and it is clean.
Not sure if there is any other AV1 encoders I could try out - or what not of settings.

Lets see if I can add a picture of the issue. Actually, I have some posted on another board - let me just refer to that, BRB. https://github.com/hvianna/audioMotion.js/discussions/107

So I do not know - I though AOM would be the better one of the two encoders so that is what I used. But found this glitch thing, although not to bad to find something I can visual compare with, being not to easy for me.

If anyone wanna know what I kinda use, here are some of my notes:

#### AOM
#!/bin/bash
outdir="AV1c36p5g240k30d10o128LN"; 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 -pix_fmt yuv420p10le \
-crf 36 -b:v 0 -cpu-used 5 -g 240 -keyint_min 30 \
-c:a libopus -b:a 128k -af "loudnorm=I=-14:LRA=11:TP=-2" -ac 2 "$output"; done; }
echo "-#-#-#-# AV1o $outdir Encoding DONE -#-#-#-#"

#### SVT
#!/bin/bash
outdir="SVTc36p7g240k30d10o128LN-720"; 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 libsvtav1 -movflags +faststart -pix_fmt yuv420p10le -svtav1-params lp=12 \
-crf 36 -b:v 0 -preset 7 -g 240 -keyint_min 30 \
-c:a libopus -b:a 128k -af "loudnorm=I=-14:LRA=11:TP=-2" -ac 2 "$output"; done; }
echo "-#-#-#-# AV1o $outdir Encoding DONE -#-#-#-#"

#### AOM 5.1t2
#!/bin/bash
outdir="AOMc36p5g240k30d10LN128-720"; 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 → $output"
ffmpeg -hide_banner -loglevel warning -y -i "$input" \
-c:v libaom-av1 -movflags +faststart -pix_fmt yuv420p10le -progress - -nostats \
-crf 36 -b:v 0 -cpu-used 5 -g 240 -keyint_min 30 -vf "scale=-1:720,crop=1280:720" \
-c:a libopus -b:a 128k -af "pan=stereo|FL=FL+0.5*FC+0.3*SL+0.3*BL|FR=FR+0.5*FC+0.3*SR+0.3*BR, loudnorm=I=-14:LRA=11:TP=-2" "$output"; done; }
echo "-#-#-#-# AV1o $outdir Encoding DONE -#-#-#-#"

#### SVT
#!/bin/bash
outdir="SVTc36p3g240k30d10o128LN-720"; 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 → $output"
ffmpeg -hide_banner -loglevel warning -y -i "$input" \
-c:v libsvtav1 -movflags +faststart -pix_fmt yuv420p10le -svtav1-params lp=10 -progress - -nostats \
-crf 36 -b:v 0 -preset 3 -g 240 -keyint_min 30 -vf "scale=-1:720,crop=1280:720" \
-c:a libopus -b:a 128k -af "pan=stereo|FL=FL+0.5*FC+0.3*SL+0.3*BL|FR=FR+0.5*FC+0.3*SR+0.3*BR, loudnorm=I=-14:LRA=11:TP=-2" "$output"; done; }
echo "-#-#-#-# AV1o $outdir Encoding DONE -#-#-#-#"

- Darkijah

[JesusGod-Pope666.Info]

Hmmm seems my version do not have any psy features - I do not know if it will help or anything. But there seems to be something that could be turned on which is not in my version apparently.
It seems that STV might already have it by default, but not AOM - I would like to try test it out.

Not sure if it would be helpful or fix the issue. But it is worth a try I guess.

[James Zern]

Hi,

On Wed, Aug 6, 2025 at 9:54 AM JesusGod-Pope666.Info <paven...@gmail.com> wrote:

Kinda newbie here:

I have this weird glitch showing up in AOM, even at 25 CFR in this movie called Nefarious in the dark shadows. It seems like it copies a table leg maybe and expands it in front of the guy walking in the shadows. I tested with SVT even on very low settings, and it is clean.

The position of the top left table leg does shift slightly in the SVT-AV1 frame, but not the libaom one. I do see a minor artifact in the libaom frame, though overall the quality looks higher.

libaom has not been tuned for one pass encoding in non-realtime modes. You might try a 2 pass encode (an encode with `-pass 1`, followed by one with `-pass 2`) to see if it helps.

 

--
You received this message because you are subscribed to the Google Groups "AV1 Discussion" group.
To unsubscribe from this group and stop receiving emails from it, send an email to av1-discuss...@aomedia.org.
To view this discussion visit https://groups.google.com/a/aomedia.org/d/msgid/av1-discuss/e3969f0f-d125-40b4-9a3d-bce7a8a38a6fn%40aomedia.org.

[JesusGod-Pope666.Info]

I do not see anything in SVT - the leg issue is not there as far as I can see. But the AOM even in 25 CFR, you can still see a slight glitch.
I have tried the 2 pass with AOM and it did nothing. Every encoding I have made with AOM in this video of Nefarious has the issue, although at 25 CFR it is very little, but still there.
SVT does not have it, even in the lowest settings I have done.

I was trying to compile the AOM with experiment setting 1, to get more options to see if I could fix it, some psy settings or something, but did not get any of the settings that ChatGPT talked about after compiling it over and over again.

So.... Yea, well if one talks about quality on AOM - this glitch just seems odd.

[JesusGod-Pope666.Info]

Anyone have any suggestions to sort it out with AOM? or wanna toy around with it, the movie is Nefarious at.... 3:22 when the guard walks out and continue in the shadows.

torsdag den 7. august 2025 kl. 21.57.58 UTC+2 skrev James Zern:

[JesusGod-Pope666.Info]

Okay as you seemed to say, after a little more eyeing the SVT results, there are some glitch as well in that - although less noticeble.

I seem to have found something that solves the glitch in the AOM encoder and trying different things now - although taking time. But it seems like the default setting is the issue - I have done less and more lag in frames and it seems the issue disappear, but use the 35 default and it comes around.
So it seems other settings of -lag-in-frames can help the issue or sort it out. Not sure why the default would have the issue. I am going to do more testing. making a 34 and 36 LIF and see if the problem is to be found or if it disappears.

[JesusGod-Pope666.Info]

So sometimes it seem to work and sometimes not - kinda weird. But it seems some of the settings make it go away on the LIF.

[James Zern]

On Sun, Aug 10, 2025 at 2:46 PM JesusGod-Pope666.Info <paven...@gmail.com> wrote:

So sometimes it seem to work and sometimes not - kinda weird. But it seems some of the settings make it go away on the LIF.

You can try using a lower `-cpu-used` value. 5 makes tradeoffs for speed with the loss of some quality. 0 will give you the best quality. Please see if the issue exists with that setting.

It's possible there is an underlying issue in the encoder. For someone to look into it, it would be helpful to file a bug (https://aomedia.googlesource.com/aom#bug-reports) with command lines and source video (if you can share it) to reproduce the problem.