Instantly Put Yourself In AI Art! FREE & Open Source!
Vložit
- čas přidán 22. 05. 2024
- Instantly Create Customized AI Images! New Open Source Stylization Method!
▼ Link(s) From Today’s Video:
Photomaker Huggingface Demo:huggingface.co/spaces/Tencent...
Photomaker Github: github.com/TencentARC/PhotoMaker
Photomaker project page: photo-maker.github.io/
► MattVidPro Discord: / discord
► Follow Me on Twitter: / mattvidpro
-------------------------------------------------
▼ Extra Links of Interest:
✩ AI LINKS MASTER LIST: www.futurepedia.io/
✩ General AI Playlist: • General MattVidPro AI ...
✩ AI I use to edit videos: www.descript.com/?lmref=nA4fDg
✩ Instagram: mattvidpro
✩ Tiktok: tiktok.com/@mattvidpro
✩ Second Channel: / @matt_pie
-------------------------------------------------
Thanks for watching Matt Video Productions! I make all sorts of videos here on CZcams! Technology, Tutorials, and Reviews! Enjoy Your stay here, and subscribe!
All Suggestions, Thoughts And Comments Are Greatly Appreciated… Because I Actually Read Them.
-------------------------------------------------
► Business Contact: MattVidProSecond@gmail.com - Věda a technologie
You have to go to User Settings, you'll find Tokens option, basically youll need to create a token.
While creating the token, you'll need to change the option from read to Write.
When creating a space for Gen Art you'll need to change the space settings instead of running on the Free CPU, you'll need to change it to the GPU option.
Normally use GPU 10A small, it's about $1.10.
YES PLEASE MAKE LOCAL INSTALLION GUIDE!!
Great video Matt. This was one that I hadn't heard of yet. Excited to try it.
A couple notes...
You can get your Hugging Face token by clicking on your profile pic, clicking on settings, and then going to "Access Tokens."
"This guy" in the beginning of the video is actually Yan LeCunn, one of the godfathers of AI. He won the Turing Award with Geoffrey Hinton for his work on ML and now runs AI at Meta. :)
Happy to see you two are watching each other ! You two are my main sources for AI and I'm pretty sure it's the case for many people here
Enjoyed your CES coverage of putting AI in everything. Keep up the good work.
Waiting for your next video 🔥
Thanks other Matt! 😂 cool to see you in the comments section! Appreciate the notes about huggingface & Yan. I thought he looked familiar! Hyped to meet up at GTC 🤙
yeah a local install tutorial would be nice! 👍
A little while back, I made an SDXL photo to represent my Synthia Nova AI singer/songwriter framework, to create a D-ID intro video for her CZcams channel if she ever gets one (which will happen as soon as Suno releases a public API.... *cough* ). I just tried this out with her photo, and damn, it works so well! So now, if/when she has her own channel, and if/when she needs to make any announcements about her channel or music... she can, and it'll still look like her 😁 We're in the AI future, everyone!
Cool
I would love to see a local install video
Yes please, Full local install video!
This is really amazing thanks for sharing. I recently developed some old films of my late grandfathers from the 1940s, some were some interesting portraits of various people he worked with. the results I have got with this tool are really very special and keep the vintage look and feel really well.
Would love to see a tutorial on how to set this up locally!
Yeah I came here for this. Hugging face doesn't give clear instructions on how to install.
This software is exactly what I imagined when I started encountering the SD shortcomings for consistent characters. I wish I had the technical knowledge to make it myself but I’m glad somebody did
This is awesome Matt! Thanks for bringing this to our attention! I hadn't heard anything about this yet. Awesome stuff! Open source ftw! 🎉
Hi Matt. Great video. Will check it out. Btw!, The black and white photo that you said was someone "long gone" is "Alan Turing". He helped crack the "Enigma code" during the second world war. He also came up with the "Turing Test" Where a human has a conversation with an AI and has to say if he/she thinks the AI is a real person. The Turing Test is considered to be the true test for AI. Can the AI fool the person into thinking it's another real person they are chatting with.
Thanks for sharing this. Looking forward to trying it out. Awesomeness. Huge step.
Thanks for creating this, spent ages playing around with it
Oh man I'm dying to see more of this! Up until now I've been using that face swap thing in discord, but this looks miles better. The jump from this time last year is crazy! I wonder where we'll be in Jan 2025
Matt, I love this! Those professional pics are you in 10 years 😜. I am not sure if I will do more of this for me or mess with friends? Good stuff. 👍
I love seeing AI advancements. I can't wait for when I can just talk to an advanced chat bot with vision to get to the result I'm looking for.
When I could upload a few images or 3d model examples and ask the AI to generate images or videos based on the characters or art styles provided. And then highlight certain areas of the result and ask for modifications or corrections if necessary. An AI that understands what it's generating could be flawless!
Love this for consistent characters - this feature alone is worth its weight in digital gold!
would def agree!
Becoming a Instagram model has never been easier or cheaper.
I would love to see a video of local installation as I've had errors following their instructions
I always find 'coder' writ instruction as if you know what they know rather than a 'How To' It's just gibberish to me :)
It’s great that this is open source. I hope they figure out how to make the clothes look consistent too. Also, I’m hoping we’ll be able to create consistent ‘backgrounds’ & consistent ‘objects’, in order to combine these things with our characters. Additionally, I wish we could pose our consistent characters in specific poses, and choose the camera angle + composition that we want. For storytelling, it would be useful if we could ‘pose multiple consistent characters together’ in an image.
All of your wishes are currently possible. Consistent Anything = Lora. Poses = ControlNet. Poses + Environment composition editing/masking.
Just use a 3-D program bro
Amazing to see this! Thank you Matt! Would love to follow how to do a local install.
The first photo guy is Yann Lecun. He's the chief AI scientist at Meta.
I actually can see this being used to generate variants of a 3D render quickly for new ideas.
Yes, this method ought to work with text-to-3D or text-to-video models.
they updated the suno ai model again the difference is pretty good not anything crazy but now the voice sounds so real with the new model if i heard it on the radio i could not tell if it was ai or not
Yeah, tried it again earlier today and it’s pretty mind-blowing. Definitely one of my favorite AI tools right now. 😊
Wow. Thanks for the hot tip! I think you are a brave man to use your own pics for the Photomaker images! Hilarious. Also - great glasses. Make you look more Vid-"PRO"!
i've been waiting for this!
You're doing so well, maybe soon there will be action figurines of you available! Oh wait... they are!
Great video, love the channel! If you have time can you answer a question. I do you get yourself in the picture narrating. Thanks
Thanks for Open Source!
You make an amazing Thanos, and I’d definitely follow you in your quest for universal conquest.
That was funny! Thanks for the vid
We need a tutorial for a local install
Seems like it might be a little cumbersome to use if it requires adding in all of the reference photos each time but if it could improve Lora training with fewer references required, that could be nice.
Hey, I just found a new voice cloning tool called easy AI voice it’s really really cool! It has over 29,000 Voice models! You need to upload recorded, audio and then it will process and you can use it. There is also a way to install it locally on your computer, which I also did.
Personally, I would love this in Pinokio! Great vid man
You and me both! Thanks!!
HF token is under your profile settings -> access tokens, obviously
Feels tons like the latent prompt blend workflow i have been editing… it doesn’t do img2img yet but the character prompt stays consistent as long as the seed is fixed on that layer in it, have to see what it does with a image input
(the workflow generates two images … and using IPAdapter and ClipVision it blends those two images together with all main aspects of the image consistent each time.
Yes please do an install guide
PLEASE MAKE A FULL LCAL INSTALL BEFORE I GO INSANE
I would LOVE a video on how to install this on a PC as I'm a total newbie :-) Heres hoping :-)
wow that's so cool, the steve jobs was on point haha
I can't wait for them to come out with better custom video upload to movie tech for open source.
What about your security of personal data? You give your photo to company and they can use it in the future.
bit of a noob here, but... how do I download this to run locally? Is it its own thing or do I put it in Autmoatic 1111 as a model?
ooh ooh Sam Altman as the main character in Manosphere
I don't know what is most scary. Thanos Matt or Lawyer Matt.
_"When life gives you lemons, become a lemon-themed ninja."_
Couldn't get it to work locally, keep giving me CUDA memory not enough. I have reduced the samplings to the lowest, but still get that error message. What is the minimum requirement? I'm using 3080Ti 12GB.
Can I install this and use it locally?
any links on colab?
is there a guide to installing this?
Would love to see these features implemented into dalle4
So much good stuff has already come out in open source these last couple of months. That pretty soon you will not have to bother if you have a decent PC.
As it looks at the moment Dall-E will rather become more crippled each day. The folks at OpenAI seem to sh***ing themselves more an more. There are a TON of features missing in Dall-E that obviously are available already - but are withheld even from paying users. Not to speak about the censoring of everything that is above first graders level. Dall-E even gives me the seed number and the gen id of each image - BUT you can't even use them to create a similar image. WTF?
what's the difference than just using roop?
i wouldn't mind a step by step guide to install this tool. :)
Error occurred when executing ApplyInstantID:'embedding' ==> I'm getting this error message, does anyone know how to fix it?
"This guy", LoL 😂
will you make a video showing how to install it?
can you do batch rendering?
Please make that local tutorial, the install instructions look like a nightmare to me
how is this different from Instant Lora?
If somebody can explain how to install it locally, would be great. I tried it but i also run into errors. Either it cant find pytorch-cuda=11.8 or if i try it another way and try to start the gradio demo; it give a error with something related to "fork"
help is appreciated.
Same thing stymied me. I rebuild the cuda stuff, but still no joy. The PyTorch stuff is an unbridled mess. This is why Python in general sucks... so may install managers (Conda/Yarn/etc...) and every time you look funny at your PyTorch-based AI app, it quits working.
ah yes, matthnos
maybe finally an open source alternative to insightface face spam?
I love it, it does have some limitations but for the most part I'd say 4 out of 5 stars. I've found you can use it successfully on 2 people of different sexes. You try to do a male and female, you might have to try it a few times but it will get it eventually however 2 people of the same sex usually it doesn't work very well at all and just blends them. I did some really cool Jedi ones. I did Courtney Love as a Jedi and Robert DeNiro as a Jedi. I did some cool ones of Millie Bobby Brown as a cyborg mom. There are some celebrities it flat does not like and will not do well. Like for example I tried to do Steven Van Zandt (Sopranos and E Street Band) as a Jedi and none of them really looked like him. If you're doing one that includes a celebrity add their name in the prompt (i.e. "Courtney Love as a Jedi in a brown robe wielding a green lightsaber"). It's definitely more good than bad plus you get some hilarious fails once in awhile.
Is there a free way to insert a face into an existing image through AI?
Thanos you looks like George Clooney 😂
Lol there's also a collabs doc can. Anyone fix so that can use in collabs it's withou Gui or interfaces so it should work
On the most technical level, it could be described as "texture-to-image".
Does this work with AMD GPU?
Still loading no results
Alright outside out that...... These face swaps are getting a lot better. Imagine the time when Faceswaps can be done with two clicks of a button perfectly.
And especially when mid journey nails it 😮😮😮😮 or pieces of software that don't even require subscriptions.
@@sophia23346 I agree with what you said, but I don't think posting this in unrelated conversations is likely to help anyone.
Can you please explain how to download the model? I have no idea what I'm doing 😂
Lol… Yann LeCun the director of AI research at Meta is just “that guy” .
Lmao I am uncultured I thought he looked familiar 😂
Apropos of nothing Matt but holy crap I just sorted your vids by oldest and god damn did you sound different af back in 2013 lmao
2:47 George Clooney
"this guy" LOL
This guy is Yann "A.I" Le Cun
I want local install please
time to make ai influencer more easier. lets make money
Thanos Vid Pro
you just have to make a generic token with write access on your huggingface account
sO ITS NOT READY FOR INSTAL IN WEBUI???????????????????????
Local installation please…..
I am getting error every time
This is the end of the world of photo re facers
Ai should replace replicate actors
Local install please
Does this mean I can finally make Gene Simmons baking brownies without having to train a LORA?? :O
(Before anyone mentions IP Adapter, I'd also like to say I already tried that with both SDXL and SD 1.5 and got poor results 😅)
Sounds like it!!!!
Auto1111 implementation when?
exactly what im sayin
Barely works keeps returning an error
the same problem
I want to be Brando's Jor-el
It’s too bad, I can’t appreciate this piece of technology because I am blind However, I love your audio AI videos like music speech and, 76 I meant sound effects🎉
More AI music coming soon!
Dude I suck ass at hugging face🤦🏿 awesome at using discord tho 😆
the man running an ai channel called yann lecun "this guy" 😂😂😂😂😂😂😂😂
Broo they took down TLDraw Draw fast 😢
Roop does the same thing but better
George Clooney?
not working
The instructions say you're supposed to write "man img" or "woman img", not just "img". So maybe if you wrote "dog img" it would have worked better.
I tried that with the specific breed and it was a tad better.
Looks like it's using Fooocus and its face-swap system
No, this is its own system