Two things: 1. considering the things you've explained, this does look like it should work on Linux pretty with only minimal changes, especially because gentle works well enough in a docker container in my experience 2. 17:40 this is because your terminal isn't wide enough for the whole line, and since iirc (I could be wrong) it literally just outputs to rewind the line to the beginning, meaning that if there *was* a line break because a line was too long, it won't rewind far enough
"[...] well-polished, nicely wrapped up piece of marketable software like something that Adobe would create". As someone whose only experience with Adobe software was Flash CS5, I'm legitimately not sure if this was intended as a joke.
Either that was a backwards joke, or he was being sarcastic in a weird way. Edit: I genuinely doubt that was a joke now after that further into the video, because I just wanna say that he is really, really wrong.
To clarify, I didn't want to insult him on that. It's just, I believe Adobe products aren't that good, but well, people have their own choice of what they use.
Very cool project. Each component seems simple enough but the result is greater than the sum of its parts. The outcome looks amazing considering how few clues you give it about emotions and stuff but it probably helps a lot that your drawings are so expressive! Also the changing pose on punctuation really helps. Great work
17:18 the reason why it may be so inefficient is because Mac laptops don’t usually have the best of the best optimization for open source projects. FFMPEG is a powerful tool, however it could be much slower running under the circumstances of CPU rendering rather than GPU. This can’t be solved on WSL because it doesn’t have the means to hardware accelerate. The only way to increase speed would likely either be hacky optimization or running Linux with a good GPU for ffmpeg to use. Really glad you released the source and I will be using it for future projects!
People who want something like this on Windows may want to look into HONK. It's a low-cost, single-developer project to create a program that people can use for vtubing avatars, and will lip sync to spoken word on the fly. Just wanted to put that out there for the people who either aren't using MacOS/Linux, or want something already pretty well polished.
Hey carykh, I've watched your videos since 2019-2021. I like your channel. Your videos have alot of effort and you animate amazingly. That's why I subscribed.
Oh my, I can't believe my immense luck. I was somehow born around the time the htwins started to upload videos out of the multiple thousands of years i could have potentially been born in. I'm literally witnessing history watching lazyhk retire, and even if I'm just water in the sea, I'm happy haha.
In the rendering step, it shouldn't take that long just to overlay a few images together. It's the PNG compression slowing down the process and since they're just intermediate files so you can switch to some lower compression image format such as bmp or jpeg at the cost of larger disk usage temporarily.
Please clarify. On Github your repository is MIT licensed which requires no attribution. Yet, you say in the video to attribute you if we use the code. Is attribution necessary for derivatives?
MIT does require attribution though... "The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software."
about 17:40 : ffmpeg is breaking each line, because your terminal is to small and the one line is broken into two lines and ffmpeg only replaces the second half, because it does not know how big your terminal is. So just fullscreen your window and it will work.
Oh I was so excited to try the thing out as I literally just recorded and edited audio for a dumb video of mine. It's unfortunate that it's MacOS only, but, like, totally understandable.
this is awesome, though I wonder if there wouldn't be any way to speed up that inefficient generation...maybe multithreading? or without the billboards, you could check which frames would be duplicates, and just skip generating, and straight copy those? the billboards could be added on with ffmpeg, as another layer
I'm gonna give you all a huge tip on getting gentle, especially on Mac. If you use curl on the REST API in the pre-built app, you can get the json file needed. No need for getting therapy from compilation errors, and no need for python virtual environment shenanigans. Just bliss.
17:45 if you increase the window size, ffmpeg will update in just one line. For some reason, when it breaks it doesn't do it. I think it is because it is showing the info and then telling the terminal to delete the last line, preparing for the next update. But since it is breaking into two lines, it just do it for the last line, leaving the second to last still visible. But if you resize your terminal it works as expected.
Now I want to try my hand at putting a nice GUI on this project.. and maybe trying my hand at the DSP necessary to do phoneme alignment myself. Sounds like a fun challenge.
@ the LKH retirement video: It's a little disappointing that you're giving up on AI, since seeing how it gets built up is more interesting to me than the end results of OpenAI's models, but I respect your decision.
the whole reason i subscribed all those years ago was because i knew there had to be a future where you finally decided to release the source code... finally, that future is today.. i have no reason to be subscribed anymore (just kidding)
Cary I have a question. Sorry I cant make it to the meetup I’m not vaccinated cause I’m a kid but if your gonna host a live stream I’ll be sure to join and also is the rocky plush gonna be available online after July 30?
i have a suggestion for a cool maybe filler, what if you made or used a pre-made auto-encounter, and made a recursive loop, where you gave it a starting image, then re-gave it into the encoder, and then so on and so first until you (hopefully) get a cool animation and a different final image?
Fortunately, it does! I actually used to run this project in Ubuntu 16.04 a couple years ago, before I switched desktops. So if anything, you could say lazykh is sort of "designed for Linux"!
Its well made and its sounds fun to use, but plz try to make a windows version for the program because im a fan of your content and I dont have a mac and I cant buy a Mac because the price is too much
Mad respect for making your code available ❤️
@Gray Man 🅥 bruh imagine advertising
@Gray Man 🅥 literally stop bruh
Ha.. we should automate the whole process...
@@_general_errorI agree
Two things:
1. considering the things you've explained, this does look like it should work on Linux pretty with only minimal changes, especially because gentle works well enough in a docker container in my experience
2. 17:40 this is because your terminal isn't wide enough for the whole line, and since iirc (I could be wrong) it literally just outputs
to rewind the line to the beginning, meaning that if there *was* a line break because a line was too long, it won't rewind far enough
Would be fun to hear someone run it on Linux (since I'm a noob) he he
"[...] well-polished, nicely wrapped up piece of marketable software like something that Adobe would create".
As someone whose only experience with Adobe software was Flash CS5, I'm legitimately not sure if this was intended as a joke.
Either that was a backwards joke, or he was being sarcastic in a weird way.
Edit: I genuinely doubt that was a joke now after that further into the video, because I just wanna say that he is really, really wrong.
To clarify, I didn't want to insult him on that. It's just, I believe Adobe products aren't that good, but well, people have their own choice of what they use.
17:40 Your terminal is not wide enough to fit the whole status line, that's why it's creating new lines.
Very cool project. Each component seems simple enough but the result is greater than the sum of its parts. The outcome looks amazing considering how few clues you give it about emotions and stuff but it probably helps a lot that your drawings are so expressive! Also the changing pose on punctuation really helps. Great work
Thanks for making this code available! Really enjoyed the lazykh channel, and really excited to see how it all worked
17:18 the reason why it may be so inefficient is because Mac laptops don’t usually have the best of the best optimization for open source projects. FFMPEG is a powerful tool, however it could be much slower running under the circumstances of CPU rendering rather than GPU. This can’t be solved on WSL because it doesn’t have the means to hardware accelerate. The only way to increase speed would likely either be hacky optimization or running Linux with a good GPU for ffmpeg to use. Really glad you released the source and I will be using it for future projects!
It isnt running ffmpeg at this point- its drawing the frames correct??
Îjśijjdjsßnxnjxksnamakækdkdkßkdkkdifkdfododdiidpœæęjfjjdhlïisj
People who want something like this on Windows may want to look into HONK. It's a low-cost, single-developer project to create a program that people can use for vtubing avatars, and will lip sync to spoken word on the fly. Just wanted to put that out there for the people who either aren't using MacOS/Linux, or want something already pretty well polished.
What if I don't want to do live stuff at all? Ever? What if I just want to make a video that I then upload?
@@MeepChangeling what's the difference between on the fly and recorded for later use? A miserable little pile of bytes, pretty much.
on windows 10 you can use the windows subsystem for liinux feature
@@HAAAAAAAAHHHHHH and 11
@@Matthew_MBG yea but 11 is shit
Very neat! I'm glad you're fulfilled with what you did on the LazyKH channel.
lazykh is better than my channel
The first video I watched from your channel was that video. This brings back so much memories
The gentle library *does* run on Linux, and Windows using the docker image, as stated by the README.
Hey carykh, I've watched your videos since 2019-2021. I like your channel. Your videos have alot of effort and you animate amazingly. That's why I subscribed.
W move
@@mathshouldntdie000epic Thanks
Gentle should run fine on Windows, as it says that it will run fine on all platforms supported by docker :)
"cat, i'm recording a video"
" *MEEEOOWWW* "
cried at this part
Rip Lazykh
Oh my, I can't believe my immense luck. I was somehow born around the time the htwins started to upload videos out of the multiple thousands of years i could have potentially been born in. I'm literally witnessing history watching lazyhk retire, and even if I'm just water in the sea, I'm happy haha.
@@wamwamwam watch the video maybe?
Very neat, and awesome of you to open source this! I'm very tempted to try using this for videos on a second channel
In the rendering step, it shouldn't take that long just to overlay a few images together. It's the PNG compression slowing down the process and since they're just intermediate files so you can switch to some lower compression image format such as bmp or jpeg at the cost of larger disk usage temporarily.
One reason to use lazykh over the other options is that it's free of charge, and the source code is available.
It's a good day when cary appears to bless us with his code
I love this! Thank youuu
Almost went away after hearing it only works on Mac but found that it works on Linux in the description. 👍
Please clarify. On Github your repository is MIT licensed which requires no attribution. Yet, you say in the video to attribute you if we use the code. Is attribution necessary for derivatives?
MIT does require attribution though... "The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software."
What an obnoxious comment
@@LandonJobeAnd that's not implying attribution.
about 17:40 : ffmpeg is breaking each line, because your terminal is to small and the one line is broken into two lines and ffmpeg only replaces the second half, because it does not know how big your terminal is. So just fullscreen your window and it will work.
cary's back with his first upload in 2022
cary upload within 5 years challenge
(impossible)
(failed)
i was just rewatching the original video yesterday and wondering if the code had already been released
what a coincidence
Oh I was so excited to try the thing out as I literally just recorded and edited audio for a dumb video of mine. It's unfortunate that it's MacOS only, but, like, totally understandable.
this is awesome, though I wonder if there wouldn't be any way to speed up that inefficient generation...maybe multithreading? or without the billboards, you could check which frames would be duplicates, and just skip generating, and straight copy those? the billboards could be added on with ffmpeg, as another layer
ooh ffmpeg interesting
You’re finally back after your long hiatus :D
macOS Ventura just released To beta profiles so you really need to update :q
Finally a video on carykh after 5 months
I'm gonna give you all a huge tip on getting gentle, especially on Mac.
If you use curl on the REST API in the pre-built app, you can get the json file needed.
No need for getting therapy from compilation errors, and no need for python virtual environment shenanigans.
Just bliss.
Could you please tell me how to use it? I'll even pay you :D
CARYKH IS BACK!
I just watched the lip-sync video earlier today and then I watched this one and I was like "what do you mean YEARS AGO?" I thought it newer lol
Yay I’ve always wanted to experiment with the program you made so thanks!
Nice that gonna be evolving.
*15:00** you can hear a nice inflection when the phonemes are enunciated "so i uploaded a video"*
The fact that there's a timer for drawing the billboards kills me. xD
12:35 i see that someone is a bit lazy
jk your projects are always crazy good and the fact that you make it public is amazing, keep it up!
a bit... lazykh
17:45 if you increase the window size, ffmpeg will update in just one line. For some reason, when it breaks it doesn't do it. I think it is because it is showing the info and then telling the terminal to delete the last line, preparing for the next update. But since it is breaking into two lines, it just do it for the last line, leaving the second to last still visible. But if you resize your terminal it works as expected.
Now I want to try my hand at putting a nice GUI on this project.. and maybe trying my hand at the DSP necessary to do phoneme alignment myself. Sounds like a fun challenge.
i have been waiting for this since the original video
I just want to say this: You are amazing.
This dude's content is like crack to me I love it so much aa
24:08 IT'S HILARIOUS HOW THE CHARACTER STARTS FROWNING AND GETS SUCKED AWAY LOLOLOL
RIP Lazykh Jan 2nd 2020- May 13th 2022. You were amazing. lol I’m doing the Muslim prayer on it.
A Cary upload… on the main channel!?!?!
@ the LKH retirement video: It's a little disappointing that you're giving up on AI, since seeing how it gets built up is more interesting to me than the end results of OpenAI's models, but I respect your decision.
You can write a Dockerfile packaging your app if you want it to run on any platform. Might do an MR if I get the motivation/time 😇
any update?
the whole reason i subscribed all those years ago was because i knew there had to be a future where you finally decided to release the source code... finally, that future is today.. i have no reason to be subscribed anymore (just kidding)
I like watching this channel Carykh you are the star to me
This is epic 😎
@Gray Man 🅥 fake 🤮
Can't wait for a Windows and Linux port of some kind
"Why are there police cars outside now?"
Such a nice human/virtual being
Cary posted :D
Finally something that's only available on mac and not windows
exactly lel
Hi Cary!
I am Alex or Alexandfriends!
But you can check out M&BA on June 17 at 9PM EST.
finally, a new upload
Wow this is just amazing
Babe look! A new CaryKH video!
next kh joke must be: cary killing hotdogs
Cary is just flower with no petals
i might actually use this. this is cool!
“WHAT THE FRICK, AAAHHH” says the thumbnail
I just watched that video today and you released this one
What a coinsidence
me too
@@BradleyGamer69420 bro that was 2 years ago
Cary I have a question. Sorry I cant make it to the meetup
I’m not vaccinated cause I’m a kid but if your gonna host
a live stream I’ll be sure to join and also is the rocky plush
gonna be available online after July 30?
New favorite RUclipsr
Well I am making an object show and I think this will help me! Also keep up the awesome work!
i have a suggestion for a cool maybe filler, what if you made or used a pre-made auto-encounter, and made a recursive loop, where you gave it a starting image, then re-gave it into the encoder, and then so on and so first until you (hopefully) get a cool animation and a different final image?
would you be merging PR's submitted to it though?
I like how people think that flower is Cary with a pink wig on.
Say Cary Kangaroo Heist for the next intro
What would be the result with other languages like german or spanish?
Great work, thanks!! ❤️
Damn, that channel somehow went under my radar and now its being retired. Thanks for the tinker project though
I literally just watched the lip sync video for the first time yesterday lol
This should run on Linux pretty well especially windows and Ubuntu
WOOO
Good job releasing it!
Does it run on Linux then? The source says it should
Fortunately, it does! I actually used to run this project in Ubuntu 16.04 a couple years ago, before I switched desktops. So if anything, you could say lazykh is sort of "designed for Linux"!
@@carykh I mean, most people probably can infer that it works on Linux too but maybe you should clarify that it works on Linux aswell
@@atravellingbleach8668 Technically it should work on most POSIX systems, which would be Cygwin or WSL on windows, though that depends
adobe creating well polished software hahah, good one!
I just realized kh is just his middle name and last name initials
YES- I needed this- TY CARY!!!!
So basically a BFDI maker
heres an idea new bfdi mouths and part that look more hand drawn and a tool for auto lip sync for them
'Like something that Adobe would create'
Hah! Good one!
Its well made and its sounds fun to use, but plz try to make a windows version for the program because im a fan of your content and I dont have a mac and I cant buy a Mac because the price is too much
wow havent seen carykh in a while that's crazy
Why is this so fun to watch
Holy crap a rare Cary sighting!
We need mor acronyms Cary Kant Hear
Hello KHary!
The thumbnail: you saying wtf you saying aaaaaa
I think a good workaround to give it win support is to give it linux support, as win 10 and later can run Linux applications
I JUST WATCHED THE VIDEO YESTERDAY NO WAY
ITS ALIVE
does it work with music too?
not sure
great job bro
Did You Use Adobe Character Animator?
If Not, Any Thoughts Of Using Adobe Character Animator?
The legend returns!