LATEST AI Advances: Dreambooth, Midjourney V4, Photorealistic Text to Image Model & Google Imagen
Vložit
- čas přidán 27. 04. 2024
- AI News - Lexica's Own Text to Image, Midjourney Stomps Competition, Wombo's Dreambooth & More!
▼ Link(s) From Today’s Video:
✩ Midjourney: www.midjourney.com/
✩ Stable Diffusion Trained on Midjourney V4: huggingface.co/prompthero/mid...
✩ Midjourney Diffusion Colab: colab.research.google.com/dri...
✩ Lexica's AI Model: / media
✩ Wombo Dreambooth: www.wombo.art/booth
✩ CrAIyon: www.craiyon.com/
✩ Playground's new pricing: playgroundai.com/pricing
✩ RunwayML: app.runwayml.com/
✩ Google AI Test Kitchen: aitestkitchen.withgoogle.com/
► MattVidPro Discord: / discord
► Follow Me on Twitter: / mattvidpro
-------------------------------------------------
▼ Extra Links of Interest:
✩ General AI Playlist: • General MattVidPro AI ...
✩ AI I use to edit videos: www.descript.com/?lmref=nA4fDg
✩ Second Channel: / @matt_pie
-------------------------------------------------
00:00 Intro
01:02 Midjourney V4
02:57 Stable Diffusion X Midjourney V4
05:05 Lexica's NEW AI Text to Image
07:43 Wombo Dreambooth
09:32 PlaygroundAI Pricing
10:03 CrAIyon Update
10:17 RunwayML Update
13:11 DALL-E 2 API
13:51 AI Test Kitchen
14:45 Outro
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
What Interested you the most? I personally am fascinated by the Lexica Model
The same, kind of crazy tech now.
RunwayML outpainting looks awesome, I've yet to get to try something like that since I refuse to support OpenAI.
This is the best video I've seen of yours and I think you work well with a partner. Plus whatever you were smoking helped I think 😂. 🤘
Whatever Jack is smoking.
Hi Matt, I finally managed to get Stable Diffusion locally installed; would you please kindly teach us how to install Dreambooth, if it is an add on, into the Stable Diffusion please
The RATE of improvement in this industry is almost as impressive (and scary) as the industry itself.
This duo is unstoppable
I want an entire video of you messing around with Midjourney V4 and comparing it to V3!
Yes please!!!
Also dall-e, stable diffusion etc., like a big comparison video. 1+ hour for sure!
@@jmm5765 Yeah, I would love to see all of the top names in the text-to-image AI space put to the test against V4!
isnt it an entire new model? you can't compare it side by side, other than by the name
@@Nico1a5 can generate same prompts and compare them.
Nice, You're the only youtuber that I manage to run into as soon as a new video comes out lol
Thats good to hear 😅
I mean, if it's ethical for both Midjourney and SD to source from unauthorized artistic content, it sure is for them to source from each other's databases.
Thanks for the Stable Diffusion Trained on Midjourney V4 tip. Getting some great results from that. Cheers.
Good stuff. Although I'm surprised you didn't cover the Painthua browser outpainting tool for local stable diffusion. I've been having stupid amounts of fun with that the last two days and it works just like the outpainting method you showed here. I believe Aitrepeneur has a video on it if you want to check it out.
@Jose Gabriel Matta Gonzales click the help button in painthua to find the setting you need to turn off to fix this.
I don't want it to stop. One day, we'll have an endless 24×7 AI news channel without repeating anything. *Addicted to AI.*
In playgroundml can u upload your own images to do out painting.?
Probably generated by AI too.
@@armondtanz No idea, mate.
You and your friend are great together😂 I hope this becomes a regular thing.
Indeed! Matt even forgot prompt lemons. :0
do you know if they will roll out invitations for AI test kitchen in EU aswell? or only us for now?
This was great, make more videos with guests!
The one with the couple in cozy hoodies; man oh man it it really off with the fabric folds. I think that was the biggest giveaway to me.
Great vid. Just enough humor added to the ai news coverage 👌
I've just installed stable diffusion on my computer, it's soo cool!
Would be cool if you did a video of how to set up each one of these programs
Matt Lemon ROCKS!!! 🤘💥
Can’t get the Midjourney Colab to work. Everything loads and then sleeps without generating anything and I have Colab pro. Have you got results? If so, then a demonstration would be useful.
Have you seen Nvidia's
eDiff-I: Text-to-Image?
It combines both DALLE and CLIP a denoising network Ai.
cascade of three diffusion models
I have heard about it! I’ll talk about it soon…
Nvidia ediff-I will definitely be a game-changer, especially when it comes to controlling if and how text appears in images...
@@MattVidPro Absolutely stunning how fast this is all moving! Looks like in a year or 2 the coherency and closeness to prompt will lead to most results being close to perfect, aesthetics is already beyond belief. Video is coming along, script writing is getting better (not yet good but much improved)... How long before I just have to ask AI for a new episode of a TV series or Movie and it can give me a new episode of Stargate SG-1? Ok, maybe that would still be 10 years away (putting aside the issues of rights) but it seems to be a realistic possibility, the vocal synthesis is already there too of course.
YT won't let me post the link. Check the White papers!
@@jonathanberry9502 Bruh if AI could infinitely generate SG-1 episodes I would just stop living, I would watch Stargate forever lmao. And if we could give the AI a prompt for the episode... /drool
AMAZING!!!!!!
INSANE!
Very good
How do I sign up for these AI beta tests?
It would be cool to make a distributed dataset and training network so that all users can benefit from the results with no owner
Did I miss something. Some of these have an aspect ratio that ISN'T 1:1 is that available now in MJ?
Not in MJ v4
aww man you never look at another mans tabs.. the way he shook his head said it all lmao
hello sir can you tell me any AI software that can convert any video into animation
RunwayMLs outpainting is better than Dall-E 2 imo plus it's free.
Wot about NVIDIAs diff ... thats AMAZING !!!!
It was so kind of you to let that homeless person sit in your house for a while and benefit from your influence.
😌😌
The Darth Vader toilets did not resist the release of my lightsaber and... I have no intention of reviewing the size of the latter - I await an update of the concept, particularly the depth of the tank and I test it again - my death star will appreciate too 😁🍻
I like craiyon, it now does big images
yeees
craiyon slays
the og
Wait wait, is this New?! OMG
oh. that's great. i got a MattVidPro AI. cheer up and have a nice day.
Hey have you heard about colossal ai? Hey offer a free open source software to train models. You can supposedly finetune stable diffusion with 8gb vram.
Please do more AI news with Jack. You're so funny together.
u should do more videos with jack
Ryan Tedder is the singer of OneRepublic
About this Lexica model : too bad none of those images are showing complete hands. Hands and fingers are still (to me at least) the weakest point in ALL AI models I could try so far.
I absolutely agree, the model still looks great though!
I did see some sample images that had hands in the Twitter thread but of course we don’t know how cherry picked the results are.
@@ShawnFumo Yes you are right. I was checking the Twitter thread a few minutes only after writing my comment here and I also did see some VERY promising results concerning hands and fingers (a huge concern for me hehe). I'm now really looking forward to the Lexica open beta 🤩
try anything v3
Careful ! Some models are carrying viruses that get installed on your pc, don't download randomly
i've seen some anime art made with AI and it blew me away. the samples here look terrible. but the other samples are amazing. i got addicted for a minute and had to stop lol
he is deff the a.i man
Jack is being held hostage
No 😳😳😳😳😳
man i like kfcw
I love AI nudes! I mean news
🤔
do not 😂😂😂
Personally I think you do best alone in your vid’s 😉 or with someone else 🤷🏼♂️
"friday"
Lemon
🍋
lemon
lemonade
Guys, how much did you smoke before filming this?
Love it!!
As good as that Lexica model is, I don't use text2img AI for photo realistic images of people! What's the point? For me it's about blending artistic genres and styles and coming up with something fresh.
You serious? Making photorealistic images from nothing is far more important and useful than generating art. We already have artists for that.
@@papawolf5970 But new artstyles and new unique content is something artists can already create. AI can't make new artstyles anyway. Photographers can't create what doesn't exist. AI can. AI can create imaginary worlds and characters that are photorealistic. Can photographers do this?
lol, your friend really doesn't know what being ethical means. Going places that guy
Awesome:)
Darth vader takes no shit!!
I think this community crashed runway xD
Jeez really?
First 😂
first reply 🤣🤣
You should put that on a resume
@@Picklemedia will do. 😅 never got to be first ever on a list and had to use the oppertunity. Check that from my bucketlist 🧤
AI is truly gonna ruin this world...wow
4:45 “Do you think it’s ethical for them to train a model like this on MidJourney’s images considering MidJourney’s its own separate company with its own model they worked really hard on?”
I think it’s unethical to train any of these models on original human artists’ art without their permission. Art was put on the internet before anyone could dream that generative AI art would be a thing. Generative AI art is (barely) technically legal because it uses a series of steps that are each legal on their own (collecting and cataloging images on the internet, producing “new” images with a proprietary algorithm). But combined, the steps create a big threat to existing artists’ work. These systems (DALL•E, MidJourney, Stable Diffusion, etc.) all exploit a loophole in current copyright law (which exists to protect artists’ work). Copyright law (and laws in general) often change too slowly to adapt to new disruptive technologies like this. I think there will be lawsuits and updates to the law in coming years to address it.
And the reply: “Yeah, isn’t the whole point of this to get better, and isn’t the whole point of getting better to learn from something else?”
“You can’t argue with that.”
Open-source and public domain/copyleft licenses exist to allow this kind of collaboration, but the creator has the right to decide if they want to allow that or not. You could argue that everything would be better if it was open, but ignoring existing laws and agreements is trodding on people’s rights. Saying it doesn’t matter is naive and disrespectful, in my opinion.
I don't expect to get paid every time someone sits on a bench that I made. I just make more things to get paid.
Sorry to tell you, but you art folks will have to let go of of those old ways.
@Rob Shuttleworth Boss, a loophole is being exploited here. A product worth billions was made off the backs of artists, and only the software engineers are getting paid. Why is that so hard for you to accept? If you choose not to get paid off your bench, that's your business. But you wouldn't be out of line if you ask for payment; because you'd be rendering a service. In any case, you can't compare serving with a bench to serving with art. Monetarily, they don't even come close. Your analogy is vastly flawed.
@@blyrax Ok, a violin then.
Cope seethe dilate
@@RobShuttleworth The analogy is irrelevant. The contribution of art to AI art generators is worth billions collectively. That's a lot of money to give to these companies for free. The "fair use" Law was made before AI art generation became a thing. It needs to be adjusted. Artists are not slaves.
I dont like to be negative but that MJv4 is kinda all hype... You would need millions of images to copy a whole AI concept like that... Maybe i need to give it another try... Myself and anyone I know who has used it have been extremely disappointed...
@ 4:50 ethical?...LOL.. ethicality is the farthest concept from a technology based on ripping off artists artwork.. I mean really??LOL no one should care about copying when every AI text to image is based on learning all of the art that has ever been created.
really hope midjourney add inpainting and outpainting
I'm glad SD got better at inpainting, cause for a while DALL∙E 2 has been the only one that's somewhat decent at it.