Photoshop AI for Architects !!
Вставка
- Опубліковано 15 лип 2024
- When I saw the new Photoshop AI "generative fill" tool I just KNEW I had to explore it and share it from an architects point of view! In this video I walk through how to get this new tool and demonstrate some possible uses for us as architects!
Timestamps:
00:00 What was made in Photoshop AI Quiz
00:50 How to Install the New Photoshop AI Generative Fill Tool
01:17 Filling an architectural rendering background with Photoshop AI
02:55 Removing Objects in an architectural Renderings with Photoshop AI Generative Fill
03:30 Adding Objects in an architectural Renderings with Photoshop AI Generative Fill
05:40 Changing Materials in an architectural Rendering with Photoshop AI Generative Fill
06:42 Adding a Window in an architectural Rendering with Photoshop AI Generative Fill
07:51 Architectural Sketching with Photoshop AI Generative Fill
08:50 Architectural Interior Rendering ideas with Photoshop AI Generative Fill
10:00 Change the Background of a Photo with Photoshop AI Generative Fill
Here are some links to all of the hardware and software I use:
My Main Revit Computer (BIMBOX) - bimbox.bimafterdark.com/
How I Record My Tutorials (Camtasia Studio) - techsmith.z6rjha.net/zVvgW
My Microphone (Blue Yeti USB) - amzn.to/3992DYy
My Studio Headphones (Sennheiser HD 600) - amzn.to/2PxTFwj
My Camera for Videos and Images (Canon EOS 80D) - amzn.to/32zbpg5
The Lens I use for these videos (50mm) - amzn.to/3cvIE8D
My Webcam for Webinars (Logitech C922X) - amzn.to/2wP3AHf
My Favorite Sketchbook (“Blank” by 30x40) - amzn.to/32yAffZ
My Favorite Sketching Pens (Sharpie “Fine”) - amzn.to/3c8r8qO
Greatest Sketching Marker of All Time (Sign Pen) - amzn.to/3ceAukN
** Some of the above links are affiliate links, meaning you don’t pay anything more but I may get a small commission for a sale… Cheers! **
Thanks for watching " Photoshop AI for Architects !! "
Absolutely insane, definitely frees up a lot of time doing rendering and object placements👍
haha and we can scrap those hard drives full of "cut out" trees and people! lol
nope is not that straight to the point how U think.. to much randomic
@@theoneandonly3nasty it’s pretty good, from the video it’s nailing the perspective too
Your guides are excellent thank you, very concise and to the point. Exactly what info I wanted to hear :)
Cheers! Thanks for tuning in and the kind words! So glad you've enjoyed me work :)
@@TheRevitKid Thank you! keep it up :)
Have been messing around with the new beta so far what I find is it's good for 3 things.
1. Extending a frame. It does a good job if you make a picture a larger frame, it can extend out the background very well to match.
2. Rendering out things from an image and filling them to match the background.
3. Adding something on top of a background or changing a selected item.
What it's not great for. Adding effects over the top of an image. Or getting fine detail in highlighted areas.
But the biggest thing that is not great at. Is creating entire backgrounds. If say you have a picture of someone against a green screen, it can take out the green screen and put in a whole background however, it is not great at perspective, nor can you then go and edit it as if it was a whole background image. It doesn't fill in space behing the subject. Still better to just import a background image and edit it that way.
So basically stable diffusion inpainting and outpainting in photoshop, great
Pretty much!? Lol
Just that it's not 'stable' yet, it will throw up random stuff and doesn't have the level of prompt control like midjourney and dall-e have. That's not to say it's anything less than great and they just hopped on the train. My biggest quarrell so far is that it feels constrained in scale and quantity e.g. if you say 'a group of construction workers', it will often generate a single person. Firefly is similarly stubborn, you can yell at it as much as you want to be minimalistic, simple, single color and it will insist on hallucinating psychedelic complex stuff :D
this is wayyy better than inpainting and outpainting...tthose features kinda work but anyone who has used them can tell you, it doesnt give you coherent good results as fast as this photoshop version. asides from fixing few faces it just really struggles doing other stuff as accurately and as good as this
@@afrosymphony8207 I get why a lot of people says this is better than stable diffusion, stable diffusion is very customizable with the right model and settings you can achieve same or even better results, but i get it photoshop ai is simpler it’s basically plug and play.
@@afrosymphony8207 cheers!
That was awesome! Thanks.
cheers! Thanks for tuning in! Glad you enjoyed
I think that once developed it will help a lot with refining rendering changes without needing to rerender a whole scene. I tried it on a project just to see what it could do. It was able to add a tv screen to a wall near perfectly. It was able to decently replicate a storefront to run to the edge of the view field. It could not come even remotely close to filling an event space with indicative seating...that will be super helpful when it can do that.
yeah… I noticed when I tried filling in a “construction site” it wasn’t great. but individual elements were awesome. I think there needs to be more context in the image itself to do things like FILL groups of objects that don’t exist…
This is nuts! 😮
Right?! 🤯
can't wait for generative fill into Adobe Premiere product :D
I’ve started to use it on architectural visuals too! However I’m only using ai on foliage / people / water etc. and not the actual architecture itself 👍 For me it’s been most beneficial on conceptual stages and pitch decks only.
For sure! Thanks for tuning in and sharing!
Excellent video!!
Cheers!
Selection based AI is the tool! I guess Autodesk could centralize everything 3D the same way
Thanks for the video
Cheers!!!
This is actually good my brother.
Thanks! 😬
That's really cool
Cheers!
Holy crap. I don't know how long it would have taken me to notice this capacity if you hadn't thrown it in front of me.
Woop!! 💪
This is a game changer !
🤘🤘 for real!!
wow!
👍
Hi. Excellent video. Very interesting content.
Do you know if it's possible to use it for making selections? For instance: select all the windows on a facade.
I tried a workaround by writing variations of: make windows red, but it didn't work.
As I find it, the content-aware tool is no good for that, unfortunately.
I know it's a beta, but I don't think it's quite as easy as people are making it seem. When you used content aware fill on the child, you left half the other person, but when you used generative fill, you fully selected them both - much easier. That tree you added changed your windows. If you unhid the tree after you modified the roof, it would no longer match, as it had part of the old roof on its layer. If you modified the brightness/contrast of the plant, the surrounding area would also change. You would have to manually mask the plant first. Also, currently all the changes are relatively low resolution, and do not look good on full resolution dslr camera images. It is still a great tool (provided it is fully integrated, and not pay-per-use), but it's not quite there just yet.
Very very picky for a beta … 😜 … sounds like you Adobe could use you on the feedback forum of the beta.
even if you selected both you would never get the same clean result as the AI
@@TheRevitKid Agreed, very picky, but I have seen about 5 reviews now, and virtually no mention of any issues. It will be interesting to see the final version.
Also thing to keep in mind, this model is running in-cloud. Meaning it is not your hardware performing the load, but Adobe servers.
@@marcusanimation100 still a really cool tool!!
This feature is nuts for making stellar memes for your group of friends!
haha so true!!! 😂
Thanks for the great video!is it possible with the new AI tool to speed up the retouching of jewelry photos? To quickly remove scratches, replace stones, and so on?
I dont see why not!?
@@TheRevitKid The problem with glare, they need to be re-drawn. It cleans scratches well.
One thing that few people talk about is that the Photoshop AI "generative fill" is low resolution, 10 24 x 10 24px. So in order to get the best results you have to do your fills in 1,000 by 1,000px square pieces and it will be sharp and look a lot better.
Yes, I did not mention that in the video … my guess is it will be higher resolution but at a premium price in the future lol
Appreciation from a Pakistani 🇵🇰 Architect ❤
Cheers!! 👍
Amazing! chur!
Cheers!! 😁
For the Add plant in the interior scene, I definitely don't think it gets a great result at that type of problem solving. I'm sure it WILL! I'm approaching this from a high detail photoreal rendering point of view with what I say next. Re: the plant, the reflection in the glass, it's trying but it's nowhere close, you mention the shadows underneath as well, I've noticed the same myself playing with it. Also, some of the images it grabs don't meld well in the scene, some better than others, but some would take less time doing the old fashioned photo-montage/photobashing route after all the adjustments you would need to blend the plant in so it doesn't look like a cardboard cutout.
For general/low detailquick and dirty, and definitely autofilling backgrounds it's awesome. you got a better result with the puppy than I usually do, so for now I think I will stick to the old ways for entourage
Yeah, it is nowhere near perfect… but SO cool that it is attempting the reflections and shadows… its only a matter of time before it gets better and better.
@@TheRevitKid The resolution is a big limitation at the moment, but that's easily going to improve, but it miiiiight be paywalled! just a prediction!
if you select over into the oringinal image some instead of at the ending you wont get those lines running up and down where the seam is
Yes, thanks! I did not overlap when I did the demo... GOod tip for others to know.
You unflinchingly Thanos'd those poor children! 😱
But seriously, thanks for sharing this huge new ability!
I have no regrets. 😏
Why's my Photoshop cc 2024 can't use generative fill tool? Tell me why?
how do i use brush select like in firefly web version? i cant select an object with brush and then use ai. how shall i do it?
the only problem with this kind of post work in AI is that you have to keep track of all these changes that aren't actually happening in the 3D model.
@@Donquijote007 no that's not happening at all.
@@Donquijote007 that time might be decades. The problem is someone or something has to draw the plan to begin with. That starts with actual surveyors, engineers, etc. This simple but complicated process would have to be done by a Drone with AI. And if it's a site that needs to be demolished first then the Drone can't do anything. Or if it's a site where there are tons of trees that block the Drones maneuverability then they still can't do anything until all trees are cut down. Yet with people, they can freely walk into the thick of the trees and measure whatever they need to.
Very nice video.
However I don't seem to be able to fill the area via AI despite doing the update
what happen if you have a render of the same project in different angle and you place a new element with this? Would the result the same?
You’d be doing it IN photoshop on the image… so… same results… somewhat.
I just started trying it out. The first thing was to change the color of a car I found in a photo online. The results were bizarre...the color mostly remained original to the photo, but the entire car itself changed to a completely different brand/shape. I made several attempts, but the results I expected - changing a white car to red - did not happen.
AI is all about learning the prompts … and understanding the way the engine works. I haven’t tried an exam Pell Ike they but I sure will! Thanks for tuning in!
My understanding is that this tool references Adobe’s image library unlike MidJourney which references the entire internet. Adobe is avoiding copyright issues and also adding a revenue stream for stock.
This makes me wonder a couple things: will generative fill become a subscription add-on, and will they open up generative fill to a user’s library?
Of course it will! haha you think this will be free forever!? Come on now...
@@TheRevitKid yea, I had zero disillusion about this. I'm 99.999% confident that using Generative Fill will result in additional fees. Right now Adobe is busy using the Beta adopters to work out the kinks, but after they do, we should expect additional fees!
i can already see some fashion applications for this, like "add a pocket" here or to turn pants into shorts, etc.
Great idea !!
Would be better view years ago. Now I’m graduated ;D
I feel you…. I can’t imagine having these tools when I was in school!!
Does this feature only work with the latest version of Photographer?
Watch the video? … it’s a beta you have install through creative cloud app…
Is there a way of making the material of a building different with generative AI?
Yes. Just like the roof example I showed !
Im working on a farm/factory photo that needs a dirt road to be replaced by asphalt road and a field of dirt be replaced with a crop, I´m having some mixed results.
You have to play with the prompts and selections! it takes some time to get used to how it wants to be selected or prompted!
Take another photo
Can you turn day into night or vice versa? Thanks
Sure. I don’t see why not
Hello friend, please, my generative fill doesnt't work, must i instal creative cloud or create a count. ? thanks
Watch the video… It is a beta in the creative cloud app
How much you use Midjourney in your renderings?
Never, actually.
can it generate a new house? maybe on an empty city place?
Yes, it can. Well, just an image of a new house. Download it and try!!
i don't know how you get those result because i tried it real quick and it gave me some garbage. maybe my prompts was wrong but well, i tried things such as "realistic tree", "kids walking"
As with all AI tools … it’s all about understanding what inputs it wants… like the shape of your selection and the terms of your prompt
I accidentally chose to hide the floating bar below the image. How can I get it back?
Go to the window menu bar and turn on “contextual task bar”
Does it work on photoshop 2021? And how can i get this ai tool?
No idea... I am using Creative CLoud so I always have the latest version... you can download the beta following the method in my video.
before generative fill content aware was the only way to remove people from a photograph? are u new to PS because clone stamp tool was the original way.
Yes, clone stamp is good, too... I am sorry I did not talk about EVERY single feature in Photoshop related to this. Content aware is the closest to what this is doing. Clone stamp only uses other parts of your image, it is not the same idea as content aware at all.
@@TheRevitKid true true, i just thought that when u said that before this was the only way to get rid of a person haha. great video btw! :)
Can you get this on ipad???
Not that I know of …. I haven’t seen it
AI took our job!
No way! AI enhanced our job 😜
i dont see a generative fill option on mine
Install the beta… I show you how in the video.
@@TheRevitKid right.. already did that. still no generate section
More tools for professionals to do same things
So true… same could be said about the current last minute photoshopping we all do lol!
It’s really sad - my dreamjob is creating 3d visualization but now it seems obsolete - the only thing I have left is graphic design and animation
It's not obsolete... it's just CHANGED from what you thought it would be... Don't undervalue your skillsets.
Twinmotion on steroids
💪💪💪
This is the end of photography as we know it.
Or a new beginning….
@@TheRevitKid Depending on usage I agree. But I do see a number of caveats.
I'm doing severe weather photography and I go out shooting lightningstorms for example.
If I go out shooting lightning & I dont catch one. I could add one in there with this feature & then posting in on social saying "Look at what I shot yesterday!".
One could easily misuse this feature. I tried it yesterday to get rid of a truck & it flatout scared me & invigorated me by just how good this works (sometimes it misses though) at the same time.
Nice video btw!
There really won't be much left for an architect to do in around 5 years time.
If you think that generating images and renderings is the value an architect adds, yes... architects will now be needed. But, that it not even CLOSE to why someone hires an architect and what value we add. In fact, this just makes THAT part of our jobs easier...
I'm not simply referring to images and renderings. For example there is software in development now that can generate a multitude of fully resolved floor plans in seconds based on the exact parameters of any given project. And look at mid journey, it can output infinite photo real concepts in seconds based on nothing more than a few well worded prompts. In my humble opinion It won't be too long until all these technologies combine and we're left with nothing much to do, at least design wise.
photoshop gpt
Haha … pretty much.
why would architects photoshop rendered image? wouldnt they just edit it and rerender it?
Why do photographers use photoshop when they can just go take another picture? We use photoshop for ALL kinds of stuff in architecture… and removing an item from a rendering in 20 seconds is 1000x faster than updating a model and re rendering
@@TheRevitKid i understand that but you are creating the scene so if you dont like something u cant change it in render, but photographer cant chop down tree
A little cheating on the kids there, because in the first preview, you didn't go all around with the selection, of course, it won't delete it perfectly. It would delete it perfectly in the previous version already, as long as you select all around.
True. But it still fills better than content aware fill. And that’s the point.
@@TheRevitKid Absolulety. I tried it yesterday for the first time, especially the new erase tool, It's on another level! Hair out of the face like it's nothing, chairs on the beach behind people... You can make a whole video about that haha.
@@islalamuerte7378 haha for real!!!
adobe .. guides! make fukin guides if U think seriously about that tool or any AI tool
Zooming in and out of your face is scary, just keep one format.
Thanks for tuning in and for the comment. Sorry you found my face so scary 🫣
what a raspy voice. it is not possible to listen. Why does the compressor impose a raspy voice? big dislike.
Are you watching the same video ?
@@TheRevitKid voice creaks
0:06 “…AI”, like a door creaking
Maybe without the effect of the compressor it would not be so unpleasant
@@ko_helet as the creator and editor of the video I am asking where in the video you hear it being “raspy” or “creek”… ? Always looking for constructive criticism … I just can’t find it