Wolf ET memes

[qrzy]

Wolfenstein ET SFX
 

 

Self made video

Edited May 1, 2023 by qrzy

[Nancy]

Nancy during loading between the maps.

 

Edited May 3, 2023 by Nancy

[picklejuice]

I made a new meme guys lol 

 

 

[qrzy]

4 hours ago, Nancy said:

Nancy during loading between the maps.

 

Everyone, unless any of you have one true battlestation predicted by id software in 1999

[qrzy]

Axis wear Balenciaga!

 

 

After looking through numerous Balenciaga meme videos using AI to generate known individuals and movie characters with style, I just had to do it! This whole video creation process was extremely painful, why?

 

At first we need to prepare voice samples we want to feed into AI. Each quickchat recording starts and ends with radio noise we need to filter out. Since those noises have exact same length for all quickchats, we can use "sox <input file> <output file> trim 0.250" to remove leading 250ms and "sox <input file> <output file> reverse trim 0.290 reverse" to remove trailing 290ms of each file. (of course these are linux/wsl console programs). A simple loop will do it for all files in no time.

 

Most AI voice cloning programs are written in python. At first I wanted to use one of them, train on my dataset and generate files locally, however due to package dependence hell, bugs in the code and odd or even impossible version requirements I finally gave up after trying and trying even the most popular applications from github.

 

AI web solutions are not better. Some sites have absurd pricing, some other lure users with numerous options, except they spawn card payment window when clicked on anything else. Some sites were promising, but failed in one or another way. For example on one site I wanted to upload 60sec wav file containing mixed voices from allies or axis, this file was 2.6MB, however this application returned an error saying wav input file size must be within 10KB-20MB, WTF?

 

Finally the only sane option was speech synthesis by elevenlabs.io after paying just only 1 usd (4.75 usd per month but I can cancel subscription). To generate randomized file containing different voice samples I have wrote this script:

 

#!/bin/bash

tempfile="temporary_ffmpeg_file.txt"
outfile="random_output.wav"

if [[ -f $tempfile ]]
then
	rm $tempfile
fi

touch $tempfile

if [[ ! -d $1 ]]
then
	echo "Directory $1 doesn't exist"
	exit 1
fi

for file in $(find $(pwd)"/"$1 | grep -P '\.wav' | sort -R | head -n 100)
do
	echo "file '"$file"'" >> $tempfile
done

yes | ffmpeg -f concat -safe 0 -i $tempfile -c copy $outfile

outlength=$(soxi -D $outfile | sed -e 's/\.[0-9]*//g')

if [[ $outlength -gt 59 ]]
then
	soxfile=$outfile".sox.wav"
	sox $outfile $soxfile trim 0 60
	mv $soxfile $outfile
fi

rm $tempfile

 

After generating new voice recordings I have used screenshots made day earlier using my laptop as ETL server and PC as client.

 

The next step was to use free trial of D-ID video generator which I fed with screenshots and generated audio. Caution! D-ID generates malformed mp4 videos which play fine but crash shotcut. To fix that I had to loop through all videos using ffmpeg command: for i in {1..10}; do ffmpeg -i $i".mp4" -c copy $i"fixd.mp4"; done

 

Those still images seen in between recordings were found on Google Images. I apologize if anyone is upset by me, using those images without asking for permission. I don't monetize my videos.

 

Background music: Thip Trong - Lightvessel

Edited May 13, 2023 by qrzy

[qrzy]

I think everyone knows original xD

[Vice86]

 

 

 

[TomekKromek]

Edited May 25, 2023 by TomekKromek

[TomekKromek]

@yaku

Edited May 25, 2023 by TomekKromek

[TomekKromek]

@Vice86

Edited May 25, 2023 by TomekKromek

[Vice86]

18 minutes ago, TomekKromek said:

@Vice86

 

 

 

 

[TomekKromek]

@yaku

@Xiahou Dun


@Vice86

 

[TomekKromek]

[SunSet]

Me and boys in Special Delivery map elevator preparing to open gates for gold runners 😉

Edited May 27, 2023 by SunSet

[Vice86]

@XeRoiX