Good to see a lot of fun updates lately. I've been using Midjourney's retexture feature on the old traditional art I made, I've been able to make really nice things with it
Also OmniGen released the model and code so you can install it locally. It's an LLM that you can give images in the prompt, like: "A woman is standing in an action pose. The woman is [image1] and the pose is [image2]." And it will follow the prompt using the images as references.
16:34 Keep in mind that the weights of SD 3.0 Large were never released, it was only ever accessible through Stability AI's API. The model that people were unhappy with was SD 3.0 Medium, which does score lower than any FLUX.1 model in the comparison (although not low enough in my personal opinion).
You didn't mention Comfy UI now being its own software now. Are you waiting for it to leave Beta before making a video on it? Also good to see Flux and the reaction to SD3 slapped Stability in the face hard enough to get them back on track. Competition is always a win.
Just an FYI, I am not a yearly sub for Midjourney but I have the $60 monthly plan (I will be upgrading later to save money) but I saw this new feature was on my page and have been using it today. 10/24/2024 But I am also in the category of having 20k+ generations and had early access to alpha website, so maybe that is a reason why I can use it without the annual plan.
Ideogram Canvas sounds like it's taking AI-powered image editing to another level! When you said "it's pretty much perfect", that really sums up how far they've progressed it. I’m curious how it compares with traditional tools like Photoshop, especially for finer detail adjustments and layering. Do you think it could fully replace them someday?
Hey bro nice vid as always, can you make a video on how to create a glif to generate a person and you upload a product photo and the glif make the person holding the product?
11:30 Most Likely they are using segmentation, it's the most reliable way to recognize objects in an image and isolate them then re-render them differently.
Would really like to see OpenAI release the image gen capability of GPT-4o - they showed it off in the preview, and it looked really good, but I haven't seen any mention of it since. At this point, I would be surprised to see it released at all.
It’s amazing to see how far you’ve come since those early tutorials, showing us how to run AI models on our own machines! But I feel like the shift to promoting paid options is moving away from what made your channel unique. Many of us initially followed you for the hands-on, practical knowledge - learning to use AI without relying on expensive platforms. While it's understandable that companies now want your platform for promotion, the true value (and impact) lies in educating people on the "truly free" options, like setting up ComfyUI, or using open-source models in a more independent way. These videos empowered viewers by giving them control, rather than leading them toward paid trials or credit-limited platforms. It’s great that you’re reaching new heights, but I hope you’ll consider revisiting that core educational content. People appreciate the skills you taught that helped them become independent in the AI space, and that’s what sets you apart. Looking forward to seeing more of that impactful content from you!
I signed up for the yearly MJ sub. The inpainting quality is good. The adherence isn't that good, I had trouble even putting glasses on an image. Lastly the censoring is insane. Kind of wish I saw this ideogram feature before I subbed to MJ :P
@@2034-SWE Good luck with that. I am a complete newcomer to AI. I took out a months subscription with Midjourney. Two weeks into my subscription, I got banned. I do not know why, I still don't know why. I e-mailed them, short answer; they do not give individual reasons to why I got banned, the ban is permanent and there are no refunds. So "Buyer beware". P.S. I did not try to, nor do I want to create NSFW images.
16:07 flux looks far better and has less anatomy issues, SD3.5 will be the next biggest thing, its still a baby who hasnt been taught what we want yet but will surely surpass flux by far
I don’t get these news about ideogram. Other models had those inpainting and outpainting features for a year or two already, so, why is it worth reviewing?
you gotta give sd3.5 LARGE a serious shot. that adherence score isnt just for show. Flux has a major problem with style and sd3.5 feels like more of a midjourney competitor with flux already being outdated. on leaderboard rankings i always voted aesthetically for the 3.5 too in the blind tests
InvokeAI has been able to do this stuff (most of it) for a while now. It's local install so you need a decent computer to handle it, but it has a very good canvas toolset.
why is the download option so stupid in canvas? it moves the screen and makes my image disappear. it should just put the bounding box on the part of the canvas i'm already staring at
The Turbo version is very good, and so much faster than Flux. I run Flux Dev locally via Forge and my 3090 cranks aways for almost a minute on some images. Stability Turbo was like 5 seconds.
The Midjourney "retexture" feature looks to me like it's just some kind of ControlNet, either edge detection or depth map. Nothing particularly new. Also keep in mind that inpainting and outpainting are the same thing: both are just "generate the missing pixels to match the existing pixels", the only difference is the position of those missing pixels. So any model that can do inpainting will intrinsically also be able to do outpainting.
Nvidia Sana. Idk how nobody's talking about this. Twice the quality of flux with a 106x speedup. 4096x4096 resolution and it uses an actual LLM instead of T5 so it understands complex prompts.
flux is nice, but stable diffusion is still better if you want a certain type of image, it is bad on understanding the prompt, but is follow the words, no obese person then input is skinny
MJ 6.1 SUX! They fucked up the fingers/toes again... after repairing in 5.2. The upscaling is an absolute joke... complete shit. Leonardo is now my go-to generator.
Too much youtube advertising!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!1 Canceling my channel subscription I prefer your own in-show sponsors. Get rid of the youtube ads. They really SUCK and it makes your video flow CRAPPY
No, typing your prompt and iterating to the result you want takes half an hour to an hour. longer than, making it from scratch for characters. And definitely longer than photobashing, for anything else. We've had thousands of image creation tools for decades none of them threatened digital artists. You want pixel control, still no replacement from doing from scratch.
@mycollegeshirt I'm talking about real art not Comercial art- no knock on that but I'm talking about real draftsmanship and painters. Things that take discipline and practice and fundamentals. This generation growing up is being programed to thing that pushing a button is all it will take. Newsflash will hit them when they're old.
it's for anyone with a yearly membership, anyone with over 10k generations, or anyone who has had a subscription for the last 12 months
Love the ideogram extend...you were looking ripped!
Good to see a lot of fun updates lately. I've been using Midjourney's retexture feature on the old traditional art I made, I've been able to make really nice things with it
Also OmniGen released the model and code so you can install it locally. It's an LLM that you can give images in the prompt, like: "A woman is standing in an action pose. The woman is [image1] and the pose is [image2]." And it will follow the prompt using the images as references.
16:34 Keep in mind that the weights of SD 3.0 Large were never released, it was only ever accessible through Stability AI's API.
The model that people were unhappy with was SD 3.0 Medium, which does score lower than any FLUX.1 model in the comparison (although not low enough in my personal opinion).
dopepics AI fixes this (AI Image Editor / Upscaler). AI art tools major upgrades
Omnigen is out. you're welcome.
LETS GOOO!!! Thx for the tip man!
WHERE BRÖTHER?
What can I say--- wait wrong place
Ohh my fucking God, it's just so good, I'm not able to run it tho, I'll try to run it locally tomorrow in the morning
Where did you try it @@sahilmeena485
You didn't mention Comfy UI now being its own software now.
Are you waiting for it to leave Beta before making a video on it?
Also good to see Flux and the reaction to SD3 slapped Stability in the face hard enough to get them back on track.
Competition is always a win.
I signed up for the beta but I plan to talk about it tomorrow. waiting for access then def going to cover it
Any word on a dall-e 4?
I'm looking forward to seeing Runway's Act One in use 🤗
12:17 this retexture will be nice to try different styles for my garden!
It's getting harder to choose which image generators are the best and most useful now. Ideogram and Leonardo are my go to at the moment though
Just an FYI, I am not a yearly sub for Midjourney but I have the $60 monthly plan (I will be upgrading later to save money) but I saw this new feature was on my page and have been using it today. 10/24/2024 But I am also in the category of having 20k+ generations and had early access to alpha website, so maybe that is a reason why I can use it without the annual plan.
Ideogram Canvas sounds like it's taking AI-powered image editing to another level! When you said "it's pretty much perfect", that really sums up how far they've progressed it. I’m curious how it compares with traditional tools like Photoshop, especially for finer detail adjustments and layering. Do you think it could fully replace them someday?
Hey bro nice vid as always, can you make a video on how to create a glif to generate a person and you upload a product photo and the glif make the person holding the product?
Ideogram the bomb 🎉
W vid
11:30 Most Likely they are using segmentation, it's the most reliable way to recognize objects in an image and isolate them then re-render them differently.
Thanks for the video Matt
Good to know!
Amazing, thanks for your UA-cam videos
For background try using a period since it you have to type something
Would really like to see OpenAI release the image gen capability of GPT-4o - they showed it off in the preview, and it looked really good, but I haven't seen any mention of it since. At this point, I would be surprised to see it released at all.
It’s amazing to see how far you’ve come since those early tutorials, showing us how to run AI models on our own machines! But I feel like the shift to promoting paid options is moving away from what made your channel unique. Many of us initially followed you for the hands-on, practical knowledge - learning to use AI without relying on expensive platforms.
While it's understandable that companies now want your platform for promotion, the true value (and impact) lies in educating people on the "truly free" options, like setting up ComfyUI, or using open-source models in a more independent way. These videos empowered viewers by giving them control, rather than leading them toward paid trials or credit-limited platforms.
It’s great that you’re reaching new heights, but I hope you’ll consider revisiting that core educational content. People appreciate the skills you taught that helped them become independent in the AI space, and that’s what sets you apart. Looking forward to seeing more of that impactful content from you!
I signed up for the yearly MJ sub. The inpainting quality is good. The adherence isn't that good, I had trouble even putting glasses on an image. Lastly the censoring is insane. Kind of wish I saw this ideogram feature before I subbed to MJ :P
Yeah ideogram is awesome. They deal with censorship in a reasonable manner
Maybe you can request a refund? I’m sure they’ll accept (80% likely)
@@2034-SWE Good luck with that. I am a complete newcomer to AI. I took out a months subscription with Midjourney. Two weeks into my subscription, I got banned. I do not know why, I still don't know why. I e-mailed them, short answer; they do not give individual reasons to why I got banned, the ban is permanent and there are no refunds. So "Buyer beware".
P.S. I did not try to, nor do I want to create NSFW images.
@@ian1957ruth refund with your card
So much great content! How do you manage to speak without any fill words? Do you have a script?
No script but I usually have outlines. I edit fill words out sometimes too. appreciate the kind words!
Can any of these AI software's let you import your images and not change the regional image in any way
Can you review runways new feature "act 1." On consistent characters?
Once I get access!
@@MattVidPro cool. It sounds like a game changer
16:07 flux looks far better and has less anatomy issues, SD3.5 will be the next biggest thing, its still a baby who hasnt been taught what we want yet but will surely surpass flux by far
You should explore the other tools in Ideogram's canvas feature.... such as the inpainting.
I don’t get these news about ideogram. Other models had those inpainting and outpainting features for a year or two already, so, why is it worth reviewing?
you gotta give sd3.5 LARGE a serious shot. that adherence score isnt just for show. Flux has a major problem with style and sd3.5 feels like more of a midjourney competitor with flux already being outdated. on leaderboard rankings i always voted aesthetically for the 3.5 too in the blind tests
Ideogram releasing the same ability to edit external images one day ahead of Midjourney is interesting
I feel bad for the cat. How do I give her cat treats? Also awesome video Matt
Nice to see Midjourney finally accepting image uploads
8:30 is there free alternative have the same level ?
Unfortunately not, although there might be some local comfy ui set ups that are able to replicate this to some degree
InvokeAI has been able to do this stuff (most of it) for a while now. It's local install so you need a decent computer to handle it, but it has a very good canvas toolset.
Where can I find a free AI canvas? Playground don't have canvas anymore
Does Leonardo still have free daily credits?
So these tools at 20 buxs or photoshop for the same with I assume AI credits?
Yep
why is the download option so stupid in canvas? it moves the screen and makes my image disappear. it should just put the bounding box on the part of the canvas i'm already staring at
1st!!!!!! I literally have discord open waiting for your videos! ❤
Also, lemon tattoos
The first canvase ai was invoke ai.
The Turbo version is very good, and so much faster than Flux. I run Flux Dev locally via Forge and my 3090 cranks aways for almost a minute on some images. Stability Turbo was like 5 seconds.
After the debacle of SD3, SAI has a long way to go to regain trust.
The Midjourney "retexture" feature looks to me like it's just some kind of ControlNet, either edge detection or depth map. Nothing particularly new.
Also keep in mind that inpainting and outpainting are the same thing: both are just "generate the missing pixels to match the existing pixels", the only difference is the position of those missing pixels. So any model that can do inpainting will intrinsically also be able to do outpainting.
Would you mind doing more self-hosted AI videos? I prefer that over endless subscriptions and the cost and limits thereof.
I think kritas diffusion plug in is super good still
Maybe it is dalle3 with lcm lora to make it faster
put a space instead of empty. works fine "usually"
The Midjourney editor is available if you've generated 10k images, not only for yearly subscribers.
If you've made 10k+ images you can also use the Midjourney editor.
Photoshop’s days are numbered. Good riddance
seems like it's been a busy week in the AI world
Hi 👋
try to get the cat to turn sideways. lol
lol 12:14 This is just controlnet depth map
Midjourney is too pricey, but it is the best no one can compete with it in realism
Nvidia Sana. Idk how nobody's talking about this. Twice the quality of flux with a 106x speedup. 4096x4096 resolution and it uses an actual LLM instead of T5 so it understands complex prompts.
Will look into it
Actually quality isn't that great from what I've seen. I was going by freschet inception distance but that's not reliable
oh my god
I've just got a heart attack
Ideogram & Flux's image generations look like cartoonish stock clip art from the 90's.
Now you’ll have to get a tattoo…
hi
OooOO
flux is nice, but stable diffusion is still better if you want a certain type of image, it is bad on understanding the prompt, but is follow the words, no obese person then input is skinny
Time for you to work out and get tattoos lol
MJ 6.1 SUX! They fucked up the fingers/toes again... after repairing in 5.2. The upscaling is an absolute joke... complete shit. Leonardo is now my go-to generator.
Poggers :o
Edit: Hi Matt
Heyo
@@MattVidPro thank you for replying matt, have a good day man 😊.
Also. Thank you for not getting tattoos. 😊
Midjourney just costs to much compared to open source free tools of the same quality
Getting big Matt, you on the roids?
As cool as this is, its nowhere even close to a replacement for Photoshop. For hobbyists perhaps. Certainly not professionals.
It's not local it's not even worth trying even if they pay me.
Too much youtube advertising!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!1
Canceling my channel subscription
I prefer your own in-show sponsors. Get rid of the youtube ads. They really SUCK and it makes your video flow CRAPPY
You covered the lamest news, yes a bunch of great things happened in AI not covered in this video.
Aaaaaa
First
Inpainting is not new
Your discord is clicky and mean lol.
my man is getting old... Just like me :(((
midjourney is the worst ai company
we are losing a generation of artist because of ai
No, typing your prompt and iterating to the result you want takes half an hour to an hour. longer than, making it from scratch for characters. And definitely longer than photobashing, for anything else. We've had thousands of image creation tools for decades none of them threatened digital artists. You want pixel control, still no replacement from doing from scratch.
@mycollegeshirt I'm talking about real art not Comercial art- no knock on that but I'm talking about real draftsmanship and painters. Things that take discipline and practice and fundamentals. This generation growing up is being programed to thing that pushing a button is all it will take. Newsflash will hit them when they're old.