Awesome, webui integration makes it so much easier than using wav2lip or thin plate spline motion model
@DoorknobHead
Жыл бұрын
Nerdy Avatar -- disable - safe - unpickle...because you don't want to have issues, man. Words to live by.
@phily708
Жыл бұрын
whats the difference? can i use it with --medvram?
@synthoelectro
Жыл бұрын
we're very close to making AI friends, they need to hear us talk to them and them talk back to us. It's just about here for all. On the local level..
@NerdyRodent
Жыл бұрын
Ikr! What a time to be alive 😀
@synthoelectro
Жыл бұрын
@@NerdyRodent imagine in the not-so-distant future. You will be able to load an AI friend, in much the same way you do by video chatting with someone. Right from your computer, no internet needed.
@MrSongib
Жыл бұрын
Congrats soon you will have a friend. xdd
@synthoelectro
Жыл бұрын
@@MrSongib and so will you and you and and you.
@nannan3347
Жыл бұрын
I soon might have to buy a second 4090 so both of my LLAMA-driven AI waifus can fight over me
@impactframes
Жыл бұрын
This is excellent the TTS is a great addition. Oh got to love cliff hangers :D 👌
@NerdyRodent
Жыл бұрын
Indeed!
@novaborngaming7826
Жыл бұрын
I love what you did with your hair. Thanks for another great video.
@NerdyRodent
Жыл бұрын
Thank you so much! 😉
@studioGZ
Жыл бұрын
Genius level play, bravo, jolly good😊👍
@NerdyRodent
Жыл бұрын
Thanks 😆
@marcdevinci893
Жыл бұрын
Wow this is awesome! At first glance, it looks as good as D-ID. EDIT Having difficulty to run it though. It would not show up at first. I had to un-install and re-install. One step that is missing is you need to install checkpoints as well but Windows Defender flagged some of the files as harmful. My mistake was to extract the .TAR files along with the .ZIP files. Those are meant to be place as is in the checkpoint folder. It works pretty well, the head seems to move a lot and the output is pretty low rez so not quite at par with D-ID. Probably a matter of time though but for now works great for a small talking avatar.
@3DLiveMusic
Жыл бұрын
hello. Install chekpoints? ca,n you explain? actualy i cant see the tab. i modified the bat file...
@3DLiveMusic
Жыл бұрын
ok. i just add the files in a folder and specifie the paths in the .bat file. work great.
@ChristDuque
Жыл бұрын
D-ID is really good... but the cost is totally insane
@ryshabh11
Жыл бұрын
Thanks
@NerdyRodent
Жыл бұрын
No problem, and thank you! 😃
@perrygolden
Жыл бұрын
Your putting out very good content in an easy to understand fashion. Have you considered exploring leveraging vector databases to extend the memory of agents?
@Dr.R.
Жыл бұрын
Thx for this Infos!
@WilfEsme
Жыл бұрын
Almost had me at your AI avatar there. I'm opting for Bluewillow AI for my avatar creator since it's free. Thank you!
@NerdyRodent
Жыл бұрын
I prefer stable diffusion as it's free :)
@BrandosLounge
Жыл бұрын
The King!
@MrDanINSANE
Жыл бұрын
First of all thank you for sharing! your videos are AWESOME! There is this problem ran again and again with some of these custom URL extensions maybe you can make a video explain how to make it work: Most custom URL won't show a the NEW TAB of whatever extension it is while it is Applied and Restart, also manually restart the .BAT file. Looking on the extensions directory the extension is already exist, but again, it won't appear on the Automatic1111 WebUI Any ideas? a quick solution? or maybe if you can make a video about this to explain how to make it work for good for others that will be great. Thanks ahead for everything you do, your content is great! 💙
@NerdyRodent
Жыл бұрын
Not sure as I can’t say I’ve faced that issue
@StringerBell
Жыл бұрын
Awesome! Is there a real self-hosted voice AI alternative like stable diffusion for images? I wonder what takes them so long with this :D
@wolfai_
Жыл бұрын
Wow, at this point I can finally stop learning and just wait for the extension to appear on the web UI one by one haha. Btw your video are cut on the end. Is that intended?
@jurandfantom
Жыл бұрын
Here are my fav talking heads in corner ! hello Nerdy! Looking better than before :) Edit: just watched video - WHAT THE FUK xD? that why it looks better!
@Byte-SizedScience
Жыл бұрын
Warning! I downloaded the checkpoint files manually since automatic install did not work for me. Unzipping one of the files - resultet in a Trojan (Trojan:Win32/Sirefef!cfg) detected by MS defender. I'm wondering why you have to disable safepickle for the install as well....seems fishy to me. Stay safe everybody!
@TransformXRED
Жыл бұрын
I'm glad that downloading the files took too long, I canceled it and I'm glad I did. I don't trust anything that requires to disable any type of of security to work. Same thing with some of the other tools that tells people to deactivate security features in windows to install some stable diffusion stuff. This doesn't worth the risks.
@cjayseven7262
Жыл бұрын
I just downloaded and scanned the files and MS Defender didn't find anything.
@swfsql
Жыл бұрын
Should run in docker or something with at least _some_ sandboxing. I do recommend AbdBarho in case of dockerized container.
@j-drum7481
Жыл бұрын
A bunch of us complained about this on the GitHub repo and they reckon the latest version has fixed it now. I assume that means they're using safetensors instead now.
@NerdyRodent
Жыл бұрын
They’ve updated now 😉
@LIMBICNATIONARTIST
Жыл бұрын
coo!
@CollinBottrell
9 ай бұрын
Having a hell of a time getting sadtalker to work on my M1 Mac
@NerdyRodent
9 ай бұрын
I went with Linux + Nvidia myself for compatibility, stability and performance!
@CollinBottrell
9 ай бұрын
I appreciate your reply @@NerdyRodent! I'm very very new to this-automatic 1111 was the first github program I've installed, it's cool but I've got a lot to learn I think. Thanks again for the reply and the great video!
@wakegary
Жыл бұрын
The avatar is f k n g back aaayyyy
@MythenAkte
Жыл бұрын
nice Video! I just set up Sadtalker and installed TTS Extension but i didn't see it in Sadtalker, how can i fix this? would love to get the TTS window in Sadtalker
@NerdyRodent
Жыл бұрын
Should be right there when you start the app!
@MythenAkte
Жыл бұрын
@@NerdyRodent when i run stable diffusion i just see sadtalker but not the TTS (i didnt see it at all)
@NerdyRodent
Жыл бұрын
@@MythenAkte yes, it’s in the app as shown and not in stable diffusion (yet!)
@ratside9485
Жыл бұрын
Nice
@NerdyRodent
Жыл бұрын
So nice!
@zoombapup
Жыл бұрын
Haha.. been looking at doing this for some lectures :) I don't like the swimming of the heads, but the mouths look alright. This has that sort of thin-plate-spline wobble a bit.
@NerdyRodent
Жыл бұрын
You can enable still mode to reduce the head movement. Personally I quite like it 😉
@nannan3347
Жыл бұрын
This is mental
@Mimeniia
Жыл бұрын
Nicu!
@BrandosLounge
Жыл бұрын
Didn't take me long to break this either, getting index out of range everytime i try to do a sample now, it was working at first but now nothing works
@NerdyRodent
Жыл бұрын
Well done! 😆 Danger is only a git pull away…
@skylerdrones777
Жыл бұрын
Your videos are great! How do you make the AI with your face over there in the corner of the video? Can you tell me which video you teach this in?
@stephanematis
Жыл бұрын
Just saw the same black cat twice. Figures, been in the Matrix all along.
@iamYork_
Жыл бұрын
cool cool... looks like much fun fun fun...
@NerdyRodent
Жыл бұрын
For sure!
@iamYork_
Жыл бұрын
@@NerdyRodent finally got around to trying it 4months later... In the AI world 4 months is 40 years so maybe this is outdated already... hahah... didn't work in latest Auto1111 for me anyway... but there can be so many factors involved in why I will just move on... haha... these AI toys break so fast...
@itaicarmeli1145
Жыл бұрын
this is like a low-level (not in a programmers' way) audio2face 🙂
@NerdyRodent
Жыл бұрын
Yup! Super simple to do 😃
@Doughy_in_the_Middle
Жыл бұрын
I just really don't like disabling safe unpickle.
@j.j.maverick9252
Жыл бұрын
It looks really good but I’ll wait for a version that passes the safety checks. Too many bad actors to play those odds now
@vincedodge321
Жыл бұрын
@@j.j.maverick9252 i agree. Not to mention that SadTalker is developed by Chinese programmers which includes TenCent AI labs.
@j-drum7481
Жыл бұрын
I think latest version no longer requires disabling safe unpickle.
@NerdyRodent
Жыл бұрын
Yup, they sorted that bit!
@viacheslavtomilov6942
Жыл бұрын
Hello! Great lesson. Could you please describe in more detail how to install TTS in the current SadTalker tab?
@NerdyRodent
Жыл бұрын
Install TTS then run the app as shown - that’s all I needed to do!
@viacheslavtomilov6942
Жыл бұрын
@@NerdyRodent i installed TTS from extansions and in folder "Installed" this is visible. But it isn't visible in SadTalker tab
@NerdyRodent
Жыл бұрын
@@viacheslavtomilov6942 yes, that’s right - it’s in the sad talker app not a1111 as yet.
@TomerGa
Жыл бұрын
Once again top-tier AI content! Your video mentioned 8GB Vram is reccoemnded, does that mean it is possible with 6GB?
@NerdyRodent
Жыл бұрын
I’ve not tried going that low, but do let me know!
@AndreasWacker
Жыл бұрын
related question: Is the workflow to make the narrator head on the lower right public, or is it a secret NR sauce? It works very well when head movement etc of the person speaking are being used.
@NerdyRodent
Жыл бұрын
It’s super secret, as shown in the video I’m afraid!
@theairchitect
Жыл бұрын
so cool! im only 6gb gpu =( sad... let me ask something? nerdy, what cam app or something do you use for real time animation that you use in left bottom footage of some of your videos?
@NerdyRodent
Жыл бұрын
Thin plate spline, though obviously sad talker for this one 😉
@theairchitect
Жыл бұрын
@@NerdyRodent i notce that this one was sad talker. do you think that this babe will run in my potato rtx 2060? =( dam rtx. why i dont by more gpu ram board rather than get a rtx board =( im stupped rs ... btw i will search for "Thin plate spline" =) always kindly nerdy! 🐀😎
@pwtucker
9 ай бұрын
Do you know if there's a way to have this effect live on a video conferencing call, like with Avatarify?
@winstoncarter8542
10 ай бұрын
What graphics card are you using? any recommendation for budget level graphics card for this purpose?
@NerdyRodent
10 ай бұрын
I’m using an old 3090, but anything Nvidia with a decent amount of VRAM (12+ GB) is good for AI
@Aiuniqreate
Жыл бұрын
Hello @Nerdy Rodent. Like everyone said, thanks for this awesome informative video. I have my Automatic1111 stable diffusion running and have installed sadtalker using the github link you shared. Made changes to .bat file as well however when I click on generate after attaching the pic and audio, I am getting Error under Generate Video section. Would appreciate if you could let me know what exactly am I missing here? 🤔
@NerdyRodent
Жыл бұрын
Tricky to say as I’ve not faced any errors as yet! Perhaps a GitHub issue with a full error log?
@Aiuniqreate
Жыл бұрын
@@NerdyRodent Hey, I did reply to you earlier but seems like it didn't go through. After fixing multiple errors, I was able to get sadtalker up and running in stable diffusion however the problem is that it's now stuck on "Face Renderer" forever and eventually failing when i upload pic and audio and try to generate a video. Could you please help me to figure out what might be the issue here?
@AaliDGr8
8 ай бұрын
i wish if in future An AI born which can do same like this video but also do hand legs and body moton hugely with realistic manners plus work with video too just like reface app in past ????
@coalbanksYQL
Жыл бұрын
Are we able to use --ref_pose and --ref_eyeblink in the Automatic 1111 UI? Would the reference video then generate a performance closer to the one you use?
@NerdyRodent
Жыл бұрын
Yup, you can just inference normally as the repo gets clones locally
@8eck
Жыл бұрын
This voice is also generated. We live in generated world already. 😁
@musumo1908
Жыл бұрын
I'm still trying to trick it to work with non human faces lol...thin plate let's you! NerdyRodent..how do you make these amazing YT videos...I want to get a little talking avatar just like you 😂...software secrets?? Thanks
@Avalon19511
Жыл бұрын
I have so many questions, I think I've worked out the answer to my previous question because it looks like in the command line your trying to point it to sadtalker which has to have been already installed, but it looks like it has a folder separate from stable diffusion???
@kaje01
Жыл бұрын
How difficult would it be to combine with open pose face to get better facial movement. Probably not far away
@synthoelectro
Жыл бұрын
Maybe... Nerdy Rodent is actually an AI...
@NerdyRodent
Жыл бұрын
As a large language model…
@synthoelectro
Жыл бұрын
@@NerdyRodent haha
@zvit
Жыл бұрын
Seems the video got cut off? :) Is there an open source way to get voices, except the painfully long tortoise-tts? (tortoise-tts-fast repo is still in maintenance mode)
@NerdyRodent
Жыл бұрын
I guess you’ve got card and other link blockers activate 😉 The link is Super Fast Voice To Voice AI! | Voice Cloning with so-vits-svc kzitem.info/news/bejne/1ZCklqCZgIFmgoY
@PrattB
Жыл бұрын
can you show us on the -disable safe unpickle - option . i still dont see the tab appear after a few tries. Hope u can do a full video on how to install . thanks
@NerdyRodent
Жыл бұрын
As per the description, you don’t need the unsafe pickle option so this full install guide is actually too full now 😆
@polystormstudio
Жыл бұрын
What did you use for the avatar explaining this video in the bottom corner?
@NerdyRodent
Жыл бұрын
Sad Talker
@polystormstudio
Жыл бұрын
@@NerdyRodent that's interesting because it looks like it's more accurate than the output you created in the demo. Is there something you did to improve the quality that wasn't shown in the tutorial?
@LaurinkoSattumaa
Жыл бұрын
Any tricks for getting this working with side profile images?
@Avalon19511
Жыл бұрын
Still trying to determine when to change the bat file before or after installing sadtalker??
@musumo1908
Жыл бұрын
Ha love this...Will it really not work with non humans?? I want a talking cat lol
@NerdyRodent
Жыл бұрын
So long as it’s got a recognisable, human face you’re golden!
@FarawayBrawlHacks
Жыл бұрын
I have not been able to get this to work, it always gives me the "out of memory" error. I do not have a lot of VRAM (GTX1060Ti), am I out of luck?
@NerdyRodent
Жыл бұрын
Yes. These days 8gb really is the minimum for gaming, etc 🫤
@banzai316
Жыл бұрын
This is one of your best recommendation. I was able to customize it to suit my needs. Great share! (YT deleted my previous comment? 😢)
@Utoko
Жыл бұрын
happens sometimes randomly yes. KZitem automod is the worst
@CultofThings
Жыл бұрын
So Sadtalker has a trojan in it called Sirefef, you really should let people know.
@NerdyRodent
Жыл бұрын
No trojans found here!
@Avalon19511
Жыл бұрын
My qyestion is do I install sadtalker then change the bat file or change the bat file then install sadtalker?
@ssserotonin1373
Жыл бұрын
After i seen a video of someone showing ai realtime masking their body and overlaying an avatar that looked hd and was a different size than their body but also erased their body entirely even when an arm or something poking out in realtime with a very realistic feather i knew its over for ar. over in its going to happen. the ingame purchases people do now for avatars in fortnite and such will make this trend probably so big it will be the first push away from smartphones. smartphones will be done by 2030 i can almost be sure theyll be looked at as blackberry type phones are today. Those ar glasses will eventually be able to let people have their own avatars. Itll be a social hype type of thing. Away from the supreme box logos, away from the instagram photo hype, into the reality / {augmented} reality of avatars and the ability to quite literally (if the avatar replacement is possible, then this is) entirely delete people from your view. In my head the feather would be difficult but more than likely i could see a function like where you mute everyone and then select base level avatars for them like a dog or a npc or a steve from minecraft lol i know it sounds wild but dont be foolish thinking meta was billions deep into the quest 2 seeing that as the pinnacle. Think metaHUMAN and that new tech. Thats where this is going theyre just needing to be established somewhere in it now thats why theyre forced to push something like meta worlds when the visual most important part and tech obviously important part are not where they know we will be soon. Gov has tech iterations ahead of civilian life. When i first watched blackmirror alot of the things were over my head and maybe a bit of the nefarious plots are just for the shock value, but the one where you could silence people around you vocally and visually really flashed in my head when i saw that tech. ANC is the newer airpods is wild. Imagine the bridge and evolution of the two mixed by then. Im just not sure how the social media would work on them cause i know they are gnna try and get the youth back on it. Maybe its relative to the avatars idk random stream of thought but wild stuff to see cause thats more than likely what this tech will be reiterated into and sold for. not a fan of corporate giants like panera releasing hand scanning biometrics and grocery stores all around me putting up facial scanners / biometric readers over all of the grocery isles in the refrigerated sections now though. Makes it not as cool to envision where we are headed
@freddybamba8025
Жыл бұрын
somehow i get stuck when downloading some part the Sadtalker never on the same but always get stuck any fix to this?
@ZeroCool22
Жыл бұрын
The generate audio from text works for Spanish language or only English?
@NerdyRodent
Жыл бұрын
I only tested the TTS in English
@GeekDynamicsLab
Жыл бұрын
It does work on spanish did couple of videos in spanish
@brendankeogh4669
Жыл бұрын
Do you know of any app that can do ‘non humans’ like a cartoon animal head
@Intunlocked
Жыл бұрын
I am trying add the Text to speech and I just don't understand your instructions. I downloaded the tts through PIP but then how do I run it as an extension?
@NerdyRodent
Жыл бұрын
Just run the app exactly as shown instead of a1111
@Intunlocked
Жыл бұрын
@@NerdyRodent I figured it out. It was me. I didn't realize that it wanted me to open a new browser with a new IP..I was doing right the whole time. Thanks for you reply though.
@EzPeasZ
Жыл бұрын
Whew, dont worry yall, we can now keep our pickles checked 😂
@NerdyRodent
Жыл бұрын
woo! yay!
@BigNewsNorman
Жыл бұрын
Can someone tell me what software is used to type in the command at 0:38 please
@TutorialesGeekReal
Жыл бұрын
How did they addressed the safe pickle situation? I've downloaded the models manually from their google drive, and added them to the folder, but I still get that on my cmd windows that the software can be dangerous :(
@TutorialesGeekReal
Жыл бұрын
Do i need to do something extra?
@NerdyRodent
Жыл бұрын
Nope, that's it - it all just works without unsafe pickles now :)
@TutorialesGeekReal
Жыл бұрын
@@NerdyRodent either I was pointing to a bad folder or the first download wasn’t the good one, I re-downloaded the models and pointed to the good folder and all in order now! It just give me an error related to ffmpeg, but the video saves on the folder, it just doesn’t appear on the webui, thanks for another great video!
@darkman237
6 ай бұрын
Is there a limit on the length of the audio file?
@zSion
6 ай бұрын
i found that there is for me, but maybe it depends how good your pc is idk. i couldnt do one minute of audio so i cut it into two parts of 30 and it worked
@xuxurlatuz
Жыл бұрын
Arggh, I push Generate button and showing an error message. I have installed ffmpeg. Any idea? Thanks
@xxxAriOrHxxx
Жыл бұрын
Same here. Have no idea what could be wrong
@victork5613
Жыл бұрын
yes, same problem. Btw its generate video in output folder, but without sound. I think its bug with ffmpeg codecs. Do you have win os? Do you have codecs installed like a K-ligh?
@xxxAriOrHxxx
Жыл бұрын
ok, i've managed to get it works. I've forgot to add ffmpeg to System variables path
@victork5613
Жыл бұрын
@@xxxAriOrHxxx yes, same )))
@beyou3015
Жыл бұрын
I have so many errors after installation
@PlanetStevens
Жыл бұрын
"Exception: Failed in loading audio2pose_checkpoint" Any idea where I can get the audio2pose checkpoint file?
@NerdyRodent
Жыл бұрын
Everything is downloaded automatically for me. Maybe restart the UI?
@PlanetStevens
Жыл бұрын
@@NerdyRodentI've reinstalled nearly 10 times. It doesn't seem to want to download the large checkpoint files. I didn't even have a "checkpoint" folder until I created one. Got "wav2lip". Now if I could only get my hands on a "audio2pose".
@NerdyRodent
Жыл бұрын
@@PlanetStevens If for whatever reason the automatic downloads fail, you can just download the files manually using any of the methods provided on the GitHub page
@kizentheslayer
Жыл бұрын
How much is the install size after adding it?
@NerdyRodent
Жыл бұрын
Around 2.5GB or so
@markdavidalcampado7784
Жыл бұрын
Iv waited for this for Sd but waiting for the better one that make it a little engaging to look.
@ZeusYTXD
Жыл бұрын
pls update the install not found new tutorial in win11
@ahminlaffet3555
Жыл бұрын
disabling safe unpickle? I mean.. really?
@NerdyRodent
Жыл бұрын
As per the video description, they’ve fixed that now 😀
@eren95666
Жыл бұрын
Why did you do the quality of your videos payed that’s sad
@SaintMatthieuSimard
Жыл бұрын
I totally lost interest in SD for all the scandals that surrounds it but that's a fascinating new addition. I'm eager for a new product that will respect copyright laws better and also be more copyrightable on its own. Also there is no limit to the disgusting things that can be done with SD. That should be addressed in the production of an ideal tool for serious content creators who don't want to trip their feet on semi-hidden laws laying on the ground.
@swfsql
Жыл бұрын
In case you're keen to doubt your own convictions on whether copyright laws are legitimate or not, I recommend the book "Against Intellectual Property". (but this doesn't relate nor address any of the other problematic or disgusting things that you referred to).
@SaintMatthieuSimard
Жыл бұрын
@@swfsql I want to make sure that I use tools that aren't landmined. SD2's training data is cleaner. I understand some people might want to do fanart, but otherwise, it's not my ultimate goal while I might want to do some honest and modest cameos here and there. I want to make sure that every tools in my toolbox are honestly earned and not to fall in a silly "Steve Akoi" kind of situation. (Dude had to actually pay for his VSTS, big news... lmao)
@shadowdemonaer
Жыл бұрын
Someone please get on an anime version
@yazmo109
Жыл бұрын
im gettng an error as well FileNotFoundError: [Errno 2] No such file or directory: 'C:\\Users\\Yaz-Desktop\\github\\stable-diffusion-webui/extensions/SadTalker/checkpoints/wav2lip.pth'
@pschoro
Жыл бұрын
disable pickles ??? that sounds fishy !! Did you look at the code of this add-on ?
@NerdyRodent
Жыл бұрын
They’ve updated now, as per the video description 😉
@elvigilantedeyoutube7214
Жыл бұрын
google colab error /content/gdrive/MyDrive/sd/stable-diffusion-webui/extensions/SadTalker/checkpoints/wav2lip.pth'
Пікірлер: 175