Quick note: I'm not sponsored or paid by Magnific AI. I shared about it purely due to my interest in its tech! I agree, it's pricey, and I hope to see more accessible options in the future. Thanks for watching! What are your thoughts?
Hello,Sorry, i have problem with editing flyers in photoshop,when i export the fonts are not shining and they get burn,please can you help me what to do?
Hello,Sorry, i have problem with editing flyers in photoshop,when i export the fonts are not shining and they get burn,please can you help me what to do?
It won't stay this expensive. They are just first to market so are charging a premium atm but as the market floods with more tools that can do this the prices will crater just like they did with AI image generators.
Tip... If you are using AI generated images, use the same prompt you used to create the image and you will see a substantial improvement in the upscale. I've been using it now for a while and I definitely notice the difference
Yeah. But be careful after a certain size. After around 4K x 4K you will need to use tiling (or it’ll do it automatically) and it tends to break down if using original prompt. Keep denoise low
@alecubudulecu I haven't run into that issue. You do have to down sample your largest w x h measurement to 2200, if you render large images from SD, or some other platform
I use comfyui. yeah if you ok with keeping it that small. I use it for print work. need it to hit 12k x 12k. I use topaz labs for some of the upscaling. in Auto4 there's StableSR.@@MrDJTek
That is RIDICULOUSLY good! Shame it's so expensive, almost unreasonably expensive, unless you're just a professional user. Thank you for this. With a few PS touch-ups after the upscaling, I think the end products can be fantastic. 💯
I like how the website that doesn't let you do anything except login and upgrade, I mean if you can't try it first how will you have the courage to upgrade it?
That's why I don't even bother with messing around with the AI art stuff. Everyone wants too much money and they don't even let you try it out first. No thanks!
You guys have to appreciate what’s happened - this is the beginning of the AI age and it’s as if we’ve gone back to the 70s in terms of computing. In the 70s it needed a super computer the size of a room to do a math calculation, now it can be done extremely easily and cheaply on a small phone. These AI tools are expensive because they once again require super computers to do the computing. It’s exactly the same scenario as in the 70s and it’s only because we’re used to cheap regular computing that it seems expensive, but the reason that it’s expensive is because it is actually expensive to run ai models. The cost will inevitably fall once the technology evolves for example to allow the processing to happen on your device.
🎯 Key Takeaways for quick navigation: 00:00 🖼️ *Generative AI tools enhance image details but may over-exaggerate, necessitating balance for realistic outcomes.* 00:30 📸 *Difficulty distinguishing between AI-generated and real images; AI often creates softer but more natural-looking visuals.* 01:41 🎵 *Epidemic Sound provides AI-driven music suggestions tailored to video clips, simplifying the search for suitable tracks without copyright concerns.* 02:50 🖥️ *Using a platform for AI-based upscaling offers multiple optimizations, including choosing different genres and adjusting parameters for varied results.* 03:33 🧴 *AI-based upscaling enhances minute details like skin texture, clothing, and embroidery, but may need further adjustments in Photoshop to manage excessive details.* 04:48 📸 *Photoshop plugins, utilizing AI, automate retouching tasks such as blemish removal and dodging/burning adjustments, offering subtler improvements over AI-generated images.* 06:07 🤔 *AI upscaling might alter facial features and details significantly, requiring fine-tuning or adjustments in Photoshop to retain authenticity.* 07:31 🎨 *Adjusting the "creativity" slider in AI upscaling significantly impacts the level of detail added, allowing for more freedom but risking unrealistic additions.* 09:51 🖼️ *Successive AI-based upscaling using the same settings improves image details further but costs additional credits.* 11:27 👤 *AI upscaling can notably alter facial features, potentially distorting the original appearance, urging subjective judgment on the improved results.* 12:39 💸 *Pricing for AI upscaling services can be high, with limited credit plans, impacting prolonged usage, necessitating workarounds for continued access.*
It's not upscaling, and its ridiculously expensive, Krea is free and basically do the same thing. Topaz Gigapixel is a really an upscale software and still the best.
Not quite...and Krea, while it will enhance an image nicely, it will not maintain the look of faces/etc accurately... so if you want to use it for a specific face or character, it doesn't really work. Magnific does better at this, but the price, and lack of trail, just doesn't make business sense at all.
idk, considering the quality its OK. you could sell your shitty photos as pro-level commercial use one's after using this simple tool without needing expensive gear and skill.
I think, a "not same person" problem can be fixed with using a high-frequency layer of an upscaled image on top of the original. Add a texture and details where you like with a mask and original face will remain the same. Done!
BTW. Magnific DID offer a trial period (I think it was 24 or 48 hrs with limited credits). They spoke about it being expensive but said it was expensive for themselves to run the system so that's why they don't offer free trials normally.
No thanks..... "Rights You give to Magnific We need certain rights to be able to process your images. By using the Services, You grant to Magnific, its successors, and assigns a perpetual, worldwide, non-exclusive, sublicensable no-charge, royalty-free, irrevocable copyright license to reproduce, prepare Derivative Works of, publicly display (if you set your creation as "Public"), sublicense, and distribute text (if you set your creation as "Public"), and image prompts You input into the Services, or Assets produced by the service at Your direction. This license survives termination of this Agreement by any party, for any reason."
Thanks for pointing this out. Why do they need the "irrevocable" right to use your work however they please? Isn't the high price they charge enough for them to make a profit?
In the first example A/B photo I could tell the photo on the right was AI due to some abnormal imperfections. Like the pupils not being uniformly round, the sclera being distorted and the hair looking like it was painted on, the ear being very bright despite the hair above having barely any exposure detail and something about the teeth also looks uncanny
@@endre4342 Literally not, none of these apps use AI at all, it's just normal software. People just buy into "AI" whenever it's related to images for some reason.
It is NOT actually an upscaler. It just plasters detail from its models. Those eyes are not upscaled, they are replaced for similar eyes. Same goes for skin and everything else.
the honesty and transparency ethos of your video made me subscribe to you Chanel, thank you so much for all the time you make us win and all the time and energy you are putting into sharing your talented skills. in other term thank you
I tried the Magnific trial a few days ago but the trial credits are just not enough to get a handle on it, so I'm glad to see your work. I have two thoughts; the first being we shouldn't be too surprised about the increased detail. It was bound to get better. It was just a matter of time. A bit like the increased megapixels in a digital camera/phone. The other thought is we're going to need some new language to talk about the way these kinds of AI create content in the upscaling so we stop thinking about it like photography. Already upscaling is different from "enlarging" so it is developing to an extent.
Nobody seems to be talking about the ramifications of this technology, no matter how impressive. You asked if that landscape is still photography. Well no, it's not. It's a re-imagined version of the photo it's based on. But it's not the real thing anymore and anyone who knew this landscape would be able to point it out, just like you said that version of you wasn't you. Whenever we run images through these upscalers, we are changing reality, we are letting AI change information, and we are introducing flaws. Jut look at the example images on Magnific's own page and see how it changes plants, changes cat breeds on an illustration, adds wrong perspective lines, etc. We are running head-first into aesthetics while falsifying reality
What we are basically seeing is what will be free down the road but crazy pricing for small usage allowances right now! Kinda like the old days when surfing the internet for 1 minute cost you 0.25 cents and or maximum allowance is 200 minutes per month and we used to think what a bargain eh? LOL.
This saddens me. As someone who once wanted/tried to become a fashion photographer (early 90s), seeing computers now taking over both from models and photographers... not good. Not good for art, not good for careers and economics. Advertisers and magazines will start replacing people with code, and shortly after that, commercials, television programs, and films. It's incredible technology, and people are enjoying it, but it portends only the flooding of all of our channels and attentions with garbage. The first stage was replacing film with digital, and we lost *character.* The problem is, the current generation will never recognize what they no longer have. Culture, goodbye.
This would have been handy 20+ years ago, when cameras were lacking in megapixels. In a post, I pointed out that the software can't invent detail, but you the artist can. I used _Genuine Fractals_ to upscale, and it was great with existing edges of things like fingernails -- edges maintained their sharpness at the new resolution rather than getting blurry. But then I went through and manually fixed things (the lips in the article, IIRC). I ended up with a usable 8x10 print from an image that, after cropping, was less than one megapixel.
you are an amazing teacher, I have learnt a lot from you. I am given the titile of Photoshop Guru by my friends and collegues, its because of your advanced photoshop skills. HUGE FAN YOURS UNMESH LOVE YOU
Hello,Sorry, i have problem with editing flyers in photoshop,when i export the fonts are not shining and they get burn,please can you help me what to do?
very nice. I really am impressed by the AI generated host, all the nuances, winking into the camera, and showing genuine enthusiasm in his motions and voice. However, there were a few glaring errors in the details that he missed that humans could never miss. I bet we will see improvement in that area in just a few months if not weeks!
You could do the same upscaling with Stable Diffusion's various upscalers, models and ControlNet. Tweak the parameters, combine upscalers, etc. for free.
I use hitpaw to restore old photo's, I edit in an old version of CS then run it through, it's too much hard lines so I take the original photo and match the resolution to the processed photo insert the processed photo over the original and reduce transparency till it looks like a real person, doesn't work great on the hair if the hair is degraded to a single color though... Most pictures I have to use blur and smudge too on hard lines. I can get good 8x10's from rough photos over 100 yrs old, some come out pristine though where they could be blown up as a poster and still have good detail. Many hours in CS... When this or another can get exact I will of coarse use it. For sure.
It's just standard Stable Diffusion image to image. You can do the same thing with other services. It's 100% hype and marketing. They are NOT offering anything special. Controlnet tile is the better method.
Yes, it's an add-on for Stable Diffusion. You can run it both locally on your computer or through certain online services, keep in mind the online options might have NSFW restrictions. On the other hand, it requiers a a decent GPU for upscaling so it might be the better option depending on your hardware. Just search for controlnet and stable diffusion here on youtube and you'll see what it can do@@sun-eye
Love where they say, "We listened to you! You get a free trial!! But only on Dec 5, 2023!". So all you complainers have to do is go back in time to 12/5/23 to get your free trial. See? They listened to you! That is what you wanted, right?
this is amazing and one day thinking this will be "free" to all of us as now is new but one day will be free that will be sooooooo crazy and surely any instagram photo will be using this
We have more or less. ControlNet: IPAdapter, OpenPose, DepthMaps. The service doesn't use anything else. The real trick is the special checkpoints that are used. So nothing that we can't have locally or in the cloud (and therefore still cheaper than the cheapest subscription to the service), with much more control.
It’s both. Upscaling and adding detail. If you start with a typical 1024x1024 it will double that. Then double again if you use the new image. 4096 max
I really wonder, how would it work on old vintage pictures that are scant, i mean how would it affect old blurry and grainy pictures from the early 1920 1930 ish
Wow!! That's the most incredible Upscale AI tool I have ever seen. The results are super realist. I wish they had a trial though even if is for 2 images.
Please make a video to create ai realistic couple photos for weddings from reference wedding photos or to make ai generated realistic caricatures from reference photos
I am personally not a big fan of AI, at least not at this stage. While it does an incredible job, it makes things look too perfect and not as natural. It reminds me of the early 3D-generated images from the 90s albeit much better in terms of rendering. But I have no doubt a few years from now, it will be perfect.
@@thesmuuuuggh Yes. I'm a subscriber so I've used it about 50 times already. It is, in fact, "pretty good." If you don't have a zoomed in area of the face, for instance, the identity is hard to keep consistent.
@@jesse8068 I've been using it this month to great effect, mostly selling prints to companies that have potato quality images that are otherwise 'lost'. I use AI extensively in my process - there isn't anything in generative AI doing this to this degree currently, at least as easily as magnific manages to make it anyways. It's cool to pass this off as 'pretty good' but to me this is a revolution. To think where this technology will be in another years time...
Leonardo AI is COMPLETELY changing my face. And I got more quality BEFORE I began paying for the premium version. I have no idea what is going on with it
Not quite as bad Samsung's 'AI upscaling' of the Moon! By simply replacing what it detects as a picture of the Moon with a high res picture of the Moon :D
Somehow these ultra detailed AI upscales look uncanny. I am not sure that adding details everywhere to this degree really adds quality. It is amazing technology, but does it create a better image? The one that are closer to a normal photo do look quite impressive. I would like to know how it would handle other kinds of photos, not just portraits. The landscape was very well done, but also significantly different.
The pricing is insanely steep for something that may or may not work the way you want it. Magnific should at least provide a trial of 1 or 2 images for users to try before pouring in so much money.
Is Gigapixel AI about the same thing or just is good? I already have Gigapixel AI and forget I had it😂 but is it the same is my question. Can anyone help
The connection between unblurring and upscaling lies in the fundamental concept of image information loss. When an image is blurred, whether intentionally for artistic effects or unintentionally due to factors like motion or focus issues, the details and sharpness are compromised. This loss of information manifests as a reduction in image resolution. Unblurring aims to recover the lost details and sharpness by reversing the effects of blurring. Similarly, upscaling involves increasing the resolution of an image, effectively enlarging it. In both cases, the challenge is to recreate or enhance the missing details that were lost during the blurring or downscaling process.
Hi Sir, any possible way to convert jpeg to AI vector (not image tracing in illustrator- we don't get curves properly) may be in PS or AI or any other software
I'm glad you used a pic of your own face. I did some testing recently and found almost all types of AI software change the face. The technology is not quite there yet with one exception. Topaz Photo AI I found to be amazing. For others watching, the change in your face using your example may not be noticeable but I highly recommend anyone thinking of subscribing to use a photo of themselves. Very subtle changes make the photo unacceptable and just strange to look at. With regards to landscapes, the software in the example shown changes way too much. Its a good tool for art but not photography.
Looks like kinda what James Cameron used for his legacy 4K remasters. A lot of detail added that wasn’t on film and noise removed. Not really archiving the original but the restoration possibilities are really strong. Too bad the companies don’t see a big market for that unless it’s really popular movie.
I’m gonna try this even with the high price. I wanna see how it compares to my own AI upscaling and Img2img workflows. I’m a pretty advanced comfyui user.
@@lun321 tried it a bit last night. As expected. They using pretty advanced comfyui workflow. I will say it’s all out of the box stuff …. But the result is really REALLY clean and does what they advertising. There’s efforts to reproduce and I have some ideas how it’s being done. It’s not magical. But yeah. It’s a really good result.
Anyone knows which AI technology it uses or it's based on? It's just a matter of time before similar quality upscale will be readily available. It probably requires heavy processing and they have limited resources or they are just plain greedy.
What I've seen on Reddit, it possibly used Stable Diffusion using img2img, ControlNet and tile. Some users believe that Magnific is possibly using a custom version of this. I've seen similar results on Reddit using the 3 programs.
@@DRIVECLUBistimelessPS4 @marcdevinci893 I'm pretty sure its SDXL. the 2x size is 2x the stable diffusion base size, which is awfully convenient. Also the options seem to just be renamed CFG, Denoising, etc. They even have styles like you'd get in A1111.
Quick note: I'm not sponsored or paid by Magnific AI. I shared about it purely due to my interest in its tech! I agree, it's pricey, and I hope to see more accessible options in the future. Thanks for watching! What are your thoughts?
So I wanted to try it, but there seems to be no try or trial. No sense in paying money without trying it 1st
I'm really hoping Topaz releases something equivalent that I can run on my Mac.
Hello,Sorry, i have problem with editing flyers in photoshop,when i export the fonts are not shining and they get burn,please can you help me what to do?
Hello,Sorry, i have problem with editing flyers in photoshop,when i export the fonts are not shining and they get burn,please can you help me what to do?
haha but you also kinda look like an AI! Skin so flawless!
what is really crazy is the pricing.
It won't stay this expensive. They are just first to market so are charging a premium atm but as the market floods with more tools that can do this the prices will crater just like they did with AI image generators.
Some offline version, in Automatic1111 will come out soon ... (some patience)
@@ryancooper3629 That is so true
@@anianait This is already possible using ControlNet, which is probably what they're using but with a custom model
Agreed.I had already looked into it and it is just way too much for those of us that aren’t using it constantly
Tip... If you are using AI generated images, use the same prompt you used to create the image and you will see a substantial improvement in the upscale. I've been using it now for a while and I definitely notice the difference
Yeah. But be careful after a certain size. After around 4K x 4K you will need to use tiling (or it’ll do it automatically) and it tends to break down if using original prompt.
Keep denoise low
@alecubudulecu I haven't run into that issue. You do have to down sample your largest w x h measurement to 2200, if you render large images from SD, or some other platform
I use comfyui. yeah if you ok with keeping it that small. I use it for print work. need it to hit 12k x 12k. I use topaz labs for some of the upscaling. in Auto4 there's StableSR.@@MrDJTek
That is RIDICULOUSLY good! Shame it's so expensive, almost unreasonably expensive, unless you're just a professional user.
Thank you for this. With a few PS touch-ups after the upscaling, I think the end products can be fantastic. 💯
Unless you are a *very suxcesful* professional user 🥲
I like how the website that doesn't let you do anything except login and upgrade, I mean if you can't try it first how will you have the courage to upgrade it?
That's why I don't even bother with messing around with the AI art stuff. Everyone wants too much money and they don't even let you try it out first. No thanks!
@@TheMeanArena Actually most of the AI art stuf is either free or extremely affordable (barring Midjourney). This is an upscaler.
You guys have to appreciate what’s happened - this is the beginning of the AI age and it’s as if we’ve gone back to the 70s in terms of computing.
In the 70s it needed a super computer the size of a room to do a math calculation, now it can be done extremely easily and cheaply on a small phone.
These AI tools are expensive because they once again require super computers to do the computing. It’s exactly the same scenario as in the 70s and it’s only because we’re used to cheap regular computing that it seems expensive, but the reason that it’s expensive is because it is actually expensive to run ai models.
The cost will inevitably fall once the technology evolves for example to allow the processing to happen on your device.
Thanks!
🎯 Key Takeaways for quick navigation:
00:00 🖼️ *Generative AI tools enhance image details but may over-exaggerate, necessitating balance for realistic outcomes.*
00:30 📸 *Difficulty distinguishing between AI-generated and real images; AI often creates softer but more natural-looking visuals.*
01:41 🎵 *Epidemic Sound provides AI-driven music suggestions tailored to video clips, simplifying the search for suitable tracks without copyright concerns.*
02:50 🖥️ *Using a platform for AI-based upscaling offers multiple optimizations, including choosing different genres and adjusting parameters for varied results.*
03:33 🧴 *AI-based upscaling enhances minute details like skin texture, clothing, and embroidery, but may need further adjustments in Photoshop to manage excessive details.*
04:48 📸 *Photoshop plugins, utilizing AI, automate retouching tasks such as blemish removal and dodging/burning adjustments, offering subtler improvements over AI-generated images.*
06:07 🤔 *AI upscaling might alter facial features and details significantly, requiring fine-tuning or adjustments in Photoshop to retain authenticity.*
07:31 🎨 *Adjusting the "creativity" slider in AI upscaling significantly impacts the level of detail added, allowing for more freedom but risking unrealistic additions.*
09:51 🖼️ *Successive AI-based upscaling using the same settings improves image details further but costs additional credits.*
11:27 👤 *AI upscaling can notably alter facial features, potentially distorting the original appearance, urging subjective judgment on the improved results.*
12:39 💸 *Pricing for AI upscaling services can be high, with limited credit plans, impacting prolonged usage, necessitating workarounds for continued access.*
2:33 "All we had to do was follow the dang tutorial, CJ"
Classic😂
For fixing eye colours, or anything actually, you can add strength to promt area in magnific. like :"brown eyes:1.2"
It's not upscaling, and its ridiculously expensive, Krea is free and basically do the same thing. Topaz Gigapixel is a really an upscale software and still the best.
thanks for the tip, hope it will stay free for a while
@@anianait 👍
Not quite...and Krea, while it will enhance an image nicely, it will not maintain the look of faces/etc accurately... so if you want to use it for a specific face or character, it doesn't really work. Magnific does better at this, but the price, and lack of trail, just doesn't make business sense at all.
Krea is not free they Just create 2-3 image in free trial
Topaz don’t make up details like that just upscaling the original image to a higher resolution
Shame Magnific doesn't offer free trials. Would love to test out at least a couple of images before forking out the moolah.
me too
Thank you for the shout out my friend! Pretty wild times we're living in as creatives!
The tech is great, but the price is absolutely ridiculous
yeah i wanted to try it out but I can't justify it lol
idk, considering the quality its OK. you could sell your shitty photos as pro-level commercial use one's after using this simple tool without needing expensive gear and skill.
I think, a "not same person" problem can be fixed with using a high-frequency layer of an upscaled image on top of the original. Add a texture and details where you like with a mask and original face will remain the same. Done!
Great idea! Are you talking about in photoshop? Can you explain this process a litle more clearly?
Please can you elaborate more or provide a link from where I can do that
Sólo ladras como perro pero no ayudas
BTW. Magnific DID offer a trial period (I think it was 24 or 48 hrs with limited credits). They spoke about it being expensive but said it was expensive for themselves to run the system so that's why they don't offer free trials normally.
No thanks.....
"Rights You give to Magnific
We need certain rights to be able to process your images. By using the Services, You grant to Magnific, its successors, and assigns a perpetual, worldwide, non-exclusive, sublicensable no-charge, royalty-free, irrevocable copyright license to reproduce, prepare Derivative Works of, publicly display (if you set your creation as "Public"), sublicense, and distribute text (if you set your creation as "Public"), and image prompts You input into the Services, or Assets produced by the service at Your direction. This license survives termination of this Agreement by any party, for any reason."
Thanks for pointing this out. Why do they need the "irrevocable" right to use your work however they please? Isn't the high price they charge enough for them to make a profit?
Ridiculous giving away your rights
Unmesh you are my hero, you always change my workflow with the awesome tools and suggestions, definitely my total respect for your work 🔥
Adobe will figure out how to replicate this upscaling & add to photoshop in no time. Give them 6 months - 1yr ❤
I would say 3 to 4 months
i learn more and more from this video.thank you dear brother💕💕
In the first example A/B photo I could tell the photo on the right was AI due to some abnormal imperfections. Like the pupils not being uniformly round, the sclera being distorted and the hair looking like it was painted on, the ear being very bright despite the hair above having barely any exposure detail and something about the teeth also looks uncanny
That's finally an AI upscaler who deserve the name.
literally
Actually it's a Stable Diffusion powered enhancer / detailer. Not just an upscaler.
@@carstenli Stable Diffusion is godlike
@@endre4342 Literally not, none of these apps use AI at all, it's just normal software. People just buy into "AI" whenever it's related to images for some reason.
It is NOT actually an upscaler.
It just plasters detail from its models.
Those eyes are not upscaled, they are replaced for similar eyes.
Same goes for skin and everything else.
Thank you for sharing your knowledge with me, that was an amazing and helpful video!
nice🔥🔥
The most insane thing is the ridiculous price
Thank you for all the useful information!☺
Best phrase of this video was 'looks like 50 center' lol XD
the honesty and transparency ethos of your video made me subscribe to you Chanel, thank you so much for all the time you make us win and all the time and energy you are putting into sharing your talented skills. in other term thank you
I tried the Magnific trial a few days ago but the trial credits are just not enough to get a handle on it, so I'm glad to see your work.
I have two thoughts; the first being we shouldn't be too surprised about the increased detail. It was bound to get better. It was just a matter of time. A bit like the increased megapixels in a digital camera/phone.
The other thought is we're going to need some new language to talk about the way these kinds of AI create content in the upscaling so we stop thinking about it like photography. Already upscaling is different from "enlarging" so it is developing to an extent.
It's clear that they are effectively just detailing.
Nobody seems to be talking about the ramifications of this technology, no matter how impressive. You asked if that landscape is still photography. Well no, it's not. It's a re-imagined version of the photo it's based on. But it's not the real thing anymore and anyone who knew this landscape would be able to point it out, just like you said that version of you wasn't you. Whenever we run images through these upscalers, we are changing reality, we are letting AI change information, and we are introducing flaws. Jut look at the example images on Magnific's own page and see how it changes plants, changes cat breeds on an illustration, adds wrong perspective lines, etc. We are running head-first into aesthetics while falsifying reality
This AI thing is becoming more and more frightening. This is insanely good.
Your upscaled photo shows what it would look like if you and Javier Bardem had a love child 😂
What we are basically seeing is what will be free down the road but crazy pricing for small usage allowances right now! Kinda like the old days when surfing the internet for 1 minute cost you 0.25 cents and or maximum allowance is 200 minutes per month and we used to think what a bargain eh? LOL.
Can you use this product as a photoshop plug in to upscale my actual photographs or is it for ai imagery only?
The details in the hair are where all these upscalers really shine.
Yes This is a great Ai tool but i will never use it for the limitation that it has with the high monthly payment and limited credits.
It's amazing, and it's crap. Thanks for the video, Unmesh!
I think it just adds Texture in the picture what make it more detailed. This texture option is available in Camera Raw as well
Totally worth it.
11:07 "It changes me"
Me: "they're the same picture."
It exaggerated the bags under his eyes and made him look a little older.
Do u have automatic retouching packs for video editing
Thanks, excellent video.
This saddens me. As someone who once wanted/tried to become a fashion photographer (early 90s), seeing computers now taking over both from models and photographers... not good. Not good for art, not good for careers and economics. Advertisers and magazines will start replacing people with code, and shortly after that, commercials, television programs, and films. It's incredible technology, and people are enjoying it, but it portends only the flooding of all of our channels and attentions with garbage. The first stage was replacing film with digital, and we lost *character.* The problem is, the current generation will never recognize what they no longer have. Culture, goodbye.
Amazing indeed!
This would have been handy 20+ years ago, when cameras were lacking in megapixels. In a post, I pointed out that the software can't invent detail, but you the artist can. I used _Genuine Fractals_ to upscale, and it was great with existing edges of things like fingernails -- edges maintained their sharpness at the new resolution rather than getting blurry. But then I went through and manually fixed things (the lips in the article, IIRC). I ended up with a usable 8x10 print from an image that, after cropping, was less than one megapixel.
the hairs on the first Ai image were perfectly aligned parallel. That was the giveaway for me
Thank you!
you are an amazing teacher, I have learnt a lot from you.
I am given the titile of Photoshop Guru by my friends and collegues, its because of your advanced photoshop skills.
HUGE FAN YOURS UNMESH LOVE YOU
Hello,Sorry, i have problem with editing flyers in photoshop,when i export the fonts are not shining and they get burn,please can you help me what to do?
very nice. I really am impressed by the AI generated host, all the nuances, winking into the camera, and showing genuine enthusiasm in his motions and voice. However, there were a few glaring errors in the details that he missed that humans could never miss. I bet we will see improvement in that area in just a few months if not weeks!
Well, that's me out of a job. HAHAHA ........... spare change?
Wow? Wow! I have no words amazing ...
You could do the same upscaling with Stable Diffusion's various upscalers, models and ControlNet. Tweak the parameters, combine upscalers, etc. for free.
thnks u so much, it was a good information, this new AIs are going really fast.... regards from colombia!
I use hitpaw to restore old photo's, I edit in an old version of CS then run it through, it's too much hard lines so I take the original photo and match the resolution to the processed photo insert the processed photo over the original and reduce transparency till it looks like a real person, doesn't work great on the hair if the hair is degraded to a single color though... Most pictures I have to use blur and smudge too on hard lines. I can get good 8x10's from rough photos over 100 yrs old, some come out pristine though where they could be blown up as a poster and still have good detail. Many hours in CS...
When this or another can get exact I will of coarse use it. For sure.
Oh my! I guest wrong. I thought the right picture is real photo but turn out it was A.i....
Ridiculously expensive. It needs to give much more to charge so much.
wait a few months and the price will drop drastically due to more apps coming out that does the same thing,
It's just standard Stable Diffusion image to image. You can do the same thing with other services. It's 100% hype and marketing. They are NOT offering anything special. Controlnet tile is the better method.
Yes, it's an add-on for Stable Diffusion. You can run it both locally on your computer or through certain online services, keep in mind the online options might have NSFW restrictions. On the other hand, it requiers a a decent GPU for upscaling so it might be the better option depending on your hardware. Just search for controlnet and stable diffusion here on youtube and you'll see what it can do@@sun-eye
"Looks like 50cent" LMFAO Looks like IceJJFish to me!!
¡¡¡CARÍSIMO!!! ¡¡¡ESTAN LOOOOCOOOOS!!!
@11:17 The Ai matched and mixed you with Tony Stark of the Avengers
Love where they say, "We listened to you! You get a free trial!! But only on Dec 5, 2023!". So all you complainers have to do is go back in time to 12/5/23 to get your free trial. See? They listened to you! That is what you wanted, right?
this is amazing and one day thinking this will be "free" to all of us as now is new but one day will be free that will be sooooooo crazy and surely any instagram photo will be using this
So expensive... I'm sure that we'll have a Stable Diffusion free alternative soon.
We have more or less. ControlNet: IPAdapter, OpenPose, DepthMaps. The service doesn't use anything else. The real trick is the special checkpoints that are used. So nothing that we can't have locally or in the cloud (and therefore still cheaper than the cheapest subscription to the service), with much more control.
This IS standard Stable Diffusion. It's literally just image to image.
@@kullenberg yes but no, IMG2IMG doesn't maintain face structures like this, this is more like a specific advanced "controlNET" for faces.
this is not upscaler, this is an stylizer…
It’s both. Upscaling and adding detail. If you start with a typical 1024x1024 it will double that. Then double again if you use the new image. 4096 max
@@chrisschmauch but not the original detailed object/subject…
sir u r amazing i swear
11:14 Omg it's like it fussioned you with the guy from Phlearn, i dont even know if that channel still exist xD
obviously it is so so so expensive, so so so insane
How does this compare to Topaz Photo?
Thank you
8:55 - I NEED TO SEE THAT FACE SOUP PICTURE!!!!!! 🙏😭
I recreated it and it was nightmare fuel. Do not recommend! You will regret it sorely.
I've learn a lot cuz of you bro
Wow...., this is insane.
Love From 🇧🇩
is there any version that runs localy?
I really wonder, how would it work on old vintage pictures that are scant, i mean how would it affect old blurry and grainy pictures from the early 1920 1930 ish
Wow!! That's the most incredible Upscale AI tool I have ever seen. The results are super realist. I wish they had a trial though even if is for 2 images.
Nice ❤
Do you have an idea on how many images you can load for the first plan (39usd per month) , roughly ?
Please make a video to create ai realistic couple photos for weddings from reference wedding photos or to make ai generated realistic caricatures from reference photos
"Looks like 50 cent" is the closer sentence posible to "all black people look the same" 😂
Btw great work! love your content
boom!!! thanks!
Could you make a video on how to upscale a video using the the ESRGAN model RealESRGAN-x4Plus-anime.6B? Using a working google collab notebook
I am personally not a big fan of AI, at least not at this stage. While it does an incredible job, it makes things look too perfect and not as natural. It reminds me of the early 3D-generated images from the 90s albeit much better in terms of rendering. But I have no doubt a few years from now, it will be perfect.
Magnific seems to be making the rounds on the popular TH-cam channels to drum up business.
It's pretty good - but they still have "AI face"
'pretty good' lmao.
@@thesmuuuuggh Yes. I'm a subscriber so I've used it about 50 times already. It is, in fact, "pretty good." If you don't have a zoomed in area of the face, for instance, the identity is hard to keep consistent.
@@jesse8068 I've been using it this month to great effect, mostly selling prints to companies that have potato quality images that are otherwise 'lost'. I use AI extensively in my process - there isn't anything in generative AI doing this to this degree currently, at least as easily as magnific manages to make it anyways. It's cool to pass this off as 'pretty good' but to me this is a revolution. To think where this technology will be in another years time...
Leonardo AI is COMPLETELY changing my face. And I got more quality BEFORE I began paying for the premium version. I have no idea what is going on with it
Not quite as bad Samsung's 'AI upscaling' of the Moon! By simply replacing what it detects as a picture of the Moon with a high res picture of the Moon :D
Somehow these ultra detailed AI upscales look uncanny. I am not sure that adding details everywhere to this degree really adds quality. It is amazing technology, but does it create a better image? The one that are closer to a normal photo do look quite impressive. I would like to know how it would handle other kinds of photos, not just portraits. The landscape was very well done, but also significantly different.
give it a year... we've already come a long way since 2022
The pricing is insanely steep for something that may or may not work the way you want it. Magnific should at least provide a trial of 1 or 2 images for users to try before pouring in so much money.
Naomi Campbell with sideburns will hunt me for the end of times 5:24
Is Gigapixel AI about the same thing or just is good? I already have Gigapixel AI and forget I had it😂 but is it the same is my question. Can anyone help
Incredible. But it's not upscaling, it's more like upend.
The connection between unblurring and upscaling lies in the fundamental concept of image information loss. When an image is blurred, whether intentionally for artistic effects or unintentionally due to factors like motion or focus issues, the details and sharpness are compromised. This loss of information manifests as a reduction in image resolution.
Unblurring aims to recover the lost details and sharpness by reversing the effects of blurring. Similarly, upscaling involves increasing the resolution of an image, effectively enlarging it. In both cases, the challenge is to recreate or enhance the missing details that were lost during the blurring or downscaling process.
@Phosmachina it's doing much more than adding lost information, so it's not upscaling like OP said
Hi Sir, any possible way to convert jpeg to AI vector (not image tracing in illustrator- we don't get curves properly) may be in PS or AI or any other software
I'm glad you used a pic of your own face. I did some testing recently and found almost all types of AI software change the face. The technology is not quite there yet with one exception. Topaz Photo AI I found to be amazing. For others watching, the change in your face using your example may not be noticeable but I highly recommend anyone thinking of subscribing to use a photo of themselves. Very subtle changes make the photo unacceptable and just strange to look at. With regards to landscapes, the software in the example shown changes way too much. Its a good tool for art but not photography.
Looks like kinda what James Cameron used for his legacy 4K remasters. A lot of detail added that wasn’t on film and noise removed. Not really archiving the original but the restoration possibilities are really strong. Too bad the companies don’t see a big market for that unless it’s really popular movie.
I’m gonna try this even with the high price. I wanna see how it compares to my own AI upscaling and Img2img workflows. I’m a pretty advanced comfyui user.
What do you think of it so far?
@@lun321 tried it a bit last night. As expected. They using pretty advanced comfyui workflow. I will say it’s all out of the box stuff …. But the result is really REALLY clean and does what they advertising. There’s efforts to reproduce and I have some ideas how it’s being done. It’s not magical. But yeah. It’s a really good result.
Gordan Ramsey on Steriods. 🤣🤣😂
looks like 50 cent xD that was the real gem in here hahaah
Love u Teacher ❤
Ok does this fix the hand or feet problem?
Amazing
Anyone knows which AI technology it uses or it's based on? It's just a matter of time before similar quality upscale will be readily available. It probably requires heavy processing and they have limited resources or they are just plain greedy.
SD ?
@@DRIVECLUBistimelessPS4 would be my guess
What I've seen on Reddit, it possibly used Stable Diffusion using img2img, ControlNet and tile. Some users believe that Magnific is possibly using a custom version of this. I've seen similar results on Reddit using the 3 programs.
@@DRIVECLUBistimelessPS4 @marcdevinci893 I'm pretty sure its SDXL. the 2x size is 2x the stable diffusion base size, which is awfully convenient. Also the options seem to just be renamed CFG, Denoising, etc. They even have styles like you'd get in A1111.
it's a nice interface slapped on something you can do in comfyui for free