Just found your channel and purchased your book on Architectural Photography. Brilliant. I have a lot of health issues (and I’m old!) and my medical team suggested photography as a hobby that would help get me outside and provide interest when stuck inside. Anyway, as a retired engineer and software developer I decided to pursue architectural photography including post processing, and your channel is a god send to an absolute beginner. Thank you for sharing your vast knowledge and expertise and I look forward to starting your book after supper. All the best from Australia.
Thank you for clarifying when it’s appropriate to remove specific elements from an architectural photograph! It will definitely speed up my post-processing by speeding up the decision process.
Stephen, I now use generative fill to add people to architectural photographs in a composition way. Especially when I’m working in public spaces or in higher education, it’s difficult to control where the people are. But I can remove people, of course, but now more importantly, I can add people, as long as they’re far enough into the frame, they look very realistic
Jack, It has been my experience in using a similar technique that the edges of the added element be very carefully controlled. The level of sharpening on the original image and the added image must coincide. Otherwise, the added element can look fake - like a flying saucer (often a pasta strainer on a string) in a 1950s sci-fi B-movie.
Agreed, what you add must blend seamlessly with the rest of the photograph. And since I’m using 100 megapixel medium format camera, I have to be careful about having the people be very deep in the scene, but still compositional. If they’re too “close“ to the camera, there artificiality is obvious.
At 17:49 This should say "Remove Tool" instead of "Generative Fill". Either tool could have been used to remove the car. In this case, I used the Remove Tool.
Thanks for your insights here. This helps rationalise what I have been doing in my architectural photography - I think calling back to the traditions of Piranesi et al makes a lot of sense. On a practical front, I find the Remove tool can do a lot of the work for me, but I also use Content Aware Fill, how do you find that versus Generative Fill?
Thanks for your note. Much as I would like to give you a concrete answer, in practice it is somewhat trial and error. The Remove Tool does most of the work in cleaning my image as you report it does for you. For larger more complicated areas, I first try Generative Fill. I start by giving it a direct prompt (more successful). If that fails, I will try without a prompt (usually less successful). If none of this produces satisfactory results, I'll try Content Aware Fill. If all fails, I will "resort" to using the Clone Stamp. But now this is rarely the case.
Steven, to remove things like that street sign with Generative Fill, you don't have to type anything in the Prompt input box. In fact, Adobe specifically says not to include words like "remove".
I think you may have gone too far with the Forum of Nerva image. The AI generated fence in the re-processed image looks completely different than the fence in the original image (that was partly obscured by the parked cars). Other than that, I like the way you used the AI tools on the other images from Rome - very effective. I've been taking a lot of inspiration from your channel. Especially fond of the references to the classical painters and engravers.
Possibly. My main concern was seeing the base of the walls and columns. The white bollards and iron fence that were eliminated were not in any way historic. Could I have reconstructed them? Well, sure. But I didn't think them that important.
The problem is, and I know Rome, that the van being parked up there is exactly what Rome is !. Rome is chaos of cars on cobbled roads amongst ancient wonders. Nobody pays the City Council!. That’s not how it works. If you edit the van out you are editing out Rome, and it’s just some random old building- if you leave it in it’s another anachronistic corner of Rome, of which there are wonderfully many in the Eternal City
The beauty of Rome is that it is open to endless artistic interpretations. Piranesi's view of the Temple of Nerva was unobstructed. With the new tools, I was finally able to follow that lead. Others in their work may certainly choose to include all the cars, trucks, and detritus for whatever reasons they have. (BTW, "Paying the City" was a joke. And the Temple of Nerva is nevertheless not some random old building.)
@@stevenbrookephotography For sure, this is a matter of individual taste and preference. I am all for eliminating obstructions that are moveable (like cars that could be there one day and drive away the next) or evidently distracting and installed without aesthetic considerations - such as trash cans, signs etc. The iron fence with white bollards, although not historic, would not fall into my own definition of "distracting objects". Even then, perhaps removing it completely would work better for me than letting AI generate a completely fake new fence. But as I said above, a matter of individual taste and preference.
I think the closest anyone has got to Piranese's views of Rome in modern era photography has been Gabriele Basilico in his Roma book where he juxtaposes his photographs against Piranese's etchings of the same buildings and venues from some 300 years ago. But even in Basilico's images, there is some "resignation" visible - the omnipresent tourists, cars and modern artefacts are there, in plain sight.
I have a commissioned shoot coming up in July of a large mixed-use apartment building. Typically, I obtain a city permit to barricade the parking directly in front of the building. A tall commercial truck or van in front can really diminish important views. Do you typically go this route or do you now depend upon AI tools to remove vehicles? My concern in this particular shoot is that because I will be so close to the building, removing large vehicles can be very time-consuming and difficult. Removal errors become quite obvious this close.
Thanks for your note, Andrew. This is a real problem for all of us. And AI is not always going to be the answer, especially if the truck/van takes up significant areas of the building. Here is what I have done in the past. If it’s a private building, I would work with the firm that contracted the photography (architect, contractor, building management company….whoever) and, as you have done, have them arrange to cordon off the critical parking/driveway areas for the time that you expect to be photographing. It’s in their best interest to do this. It is easier to do this with a morning shot. If it’s a public area it's clearly harder to arrange. And it may force you to photograph at a less-than-optimum time. Also check to see if a weekend session precludes the vehicles being there. Absent any help, I have blocked driveways, parking areas, etc myself with yellow tape and have had my assistants there to direct traffic away from the building. People complain; let them. Here is a last resort, which I have also used. It helps if it’s an axial shot. I take the photograph with the vehicle in the way. If possible, I then go between the vehicle and the building and shoot an additional view of what is being blocked at exactly the same height as the overall photo, with the same lens and the same exposure settings. Then, I very carefully layer the two images. If the light conditions are identical (e.g., the van/truck is not casting a shadow) this can actually work. The result will probably be better than trying to use AI to recreate what’s behind the van.
@@stevenbrookephotography Great thoughts, Steven. One downside to barricading parking is people ignore the signs! Your last idea is interesting. I haven't tried that approach for subjects of this nature. Will have to consider that.
LOL They usually don't plow through yellow tape. And having an assistant can help. Here is a tip: if you can get to them before they turn off their engines, you can usually get them to park elsewhere. Once the engine is off, the inertia is apparently too much for them. Good luck with this, Andrew. We are rooting for you!
So as a professional photographer how do you feel about relinquishing your property rights to adobe when you agreed to the new TOS? Which also includes blanket access and domain to all the videos, and any other work that you have used an adobe product on.
I would say that zooming up on closer inspection these tools leave a lot to be desired - strange artifacts and often unconvincing removal - looking at images not zoomed in might look ok but zoom in 100% and things often fall apart
Yes. Close inspection of the final result is certainly necessary. In many cases, it takes more than one pass with either tool to achieve a totally believable result. Generative Fill, in particular, can often create it's own structures and requires careful inspection following the manipulation. That said, I believe that these tools will continue to evolve and that the results will correspondingly improve.
I agree. In my work documenting Miami's endangered historic neighborhoods, I make a point of not removing anything. All the encroaching high rises, out-of-scale development, and other urban detritus are part of the story. But this project is more photojournalism than work for architects or designers, documenting new residences or commercial buildings.
I'm done with Adobe Stuff. I rather learn other tools and learn to complement the AI with stable diffusion. Painful Process but must be done. In the meantime So far I've have not find something that I could not do with other tools like GIMP or Davinci Resolve. (Yes, you can Edit pictures, not just edit video). Resolve Studio has a tool for wire removal and other for object removal that does exactly the same. Meaning is content aware.
Just found your channel and purchased your book on Architectural Photography. Brilliant. I have a lot of health issues (and I’m old!) and my medical team suggested photography as a hobby that would help get me outside and provide interest when stuck inside. Anyway, as a retired engineer and software developer I decided to pursue architectural photography including post processing, and your channel is a god send to an absolute beginner. Thank you for sharing your vast knowledge and expertise and I look forward to starting your book after supper. All the best from Australia.
Good luck with your new adventure! I am gratified to know that the videos will be of help to you.
You are the man. thanks Steven
You're very welcome!!
Excellent presentation as always.
I certainly learned a lot on using these AI tools.
Thank you for clarifying when it’s appropriate to remove specific elements from an architectural photograph! It will definitely speed up my post-processing by speeding up the decision process.
Stephen, I now use generative fill to add people to architectural photographs in a composition way. Especially when I’m working in public spaces or in higher education, it’s difficult to control where the people are. But I can remove people, of course, but now more importantly, I can add people, as long as they’re far enough into the frame, they look very realistic
Jack, It has been my experience in using a similar technique that the edges of the added element be very carefully controlled. The level of sharpening on the original image and the added image must coincide. Otherwise, the added element can look fake - like a flying saucer (often a pasta strainer on a string) in a 1950s sci-fi B-movie.
Agreed, what you add must blend seamlessly with the rest of the photograph. And since I’m using 100 megapixel medium format camera, I have to be careful about having the people be very deep in the scene, but still compositional. If they’re too “close“ to the camera, there artificiality is obvious.
At 17:49 This should say "Remove Tool" instead of "Generative Fill". Either tool could have been used to remove the car. In this case, I used the Remove Tool.
These are my precise thoughts about the AI tools for architectural photography, and you've expressed the ideas perfectly. Thank you Steven.
Love your videos sir. You're an amazing teacher. Never stop uploading these videos!😀🙏🏼
Thank you!
fully agree to that.
06:48 Did generative fill replace the window behind the power transformer on the pole or did you composite that window in from somewhere else?
That was taken from an existing window.
Thanks for your insights here. This helps rationalise what I have been doing in my architectural photography - I think calling back to the traditions of Piranesi et al makes a lot of sense. On a practical front, I find the Remove tool can do a lot of the work for me, but I also use Content Aware Fill, how do you find that versus Generative Fill?
Thanks for your note. Much as I would like to give you a concrete answer, in practice it is somewhat trial and error. The Remove Tool does most of the work in cleaning my image as you report it does for you. For larger more complicated areas, I first try Generative Fill. I start by giving it a direct prompt (more successful). If that fails, I will try without a prompt (usually less successful). If none of this produces satisfactory results, I'll try Content Aware Fill. If all fails, I will "resort" to using the Clone Stamp. But now this is rarely the case.
Steven, to remove things like that street sign with Generative Fill, you don't have to type anything in the Prompt input box. In fact, Adobe specifically says not to include words like "remove".
I've done it both ways. At times 'no prompt' works; other times it doesn't.
I think you may have gone too far with the Forum of Nerva image. The AI generated fence in the re-processed image looks completely different than the fence in the original image (that was partly obscured by the parked cars). Other than that, I like the way you used the AI tools on the other images from Rome - very effective. I've been taking a lot of inspiration from your channel. Especially fond of the references to the classical painters and engravers.
Possibly. My main concern was seeing the base of the walls and columns. The white bollards and iron fence that were eliminated were not in any way historic. Could I have reconstructed them? Well, sure. But I didn't think them that important.
The problem is, and I know Rome, that the van being parked up there is exactly what Rome is !. Rome is chaos of cars on cobbled roads amongst ancient wonders. Nobody pays the City Council!. That’s not how it works. If you edit the van out you are editing out Rome, and it’s just some random old building- if you leave it in it’s another anachronistic corner of Rome, of which there are wonderfully many in the Eternal City
The beauty of Rome is that it is open to endless artistic interpretations. Piranesi's view of the Temple of Nerva was unobstructed. With the new tools, I was finally able to follow that lead. Others in their work may certainly choose to include all the cars, trucks, and detritus for whatever reasons they have. (BTW, "Paying the City" was a joke. And the Temple of Nerva is nevertheless not some random old building.)
@@stevenbrookephotography For sure, this is a matter of individual taste and preference. I am all for eliminating obstructions that are moveable (like cars that could be there one day and drive away the next) or evidently distracting and installed without aesthetic considerations - such as trash cans, signs etc. The iron fence with white bollards, although not historic, would not fall into my own definition of "distracting objects". Even then, perhaps removing it completely would work better for me than letting AI generate a completely fake new fence. But as I said above, a matter of individual taste and preference.
Thanks for taking the time to join in. Sure, the AI fence could ;possibly go.
Dirty sensor lol great video
I think the closest anyone has got to Piranese's views of Rome in modern era photography has been Gabriele Basilico in his Roma book where he juxtaposes his photographs against Piranese's etchings of the same buildings and venues from some 300 years ago. But even in Basilico's images, there is some "resignation" visible - the omnipresent tourists, cars and modern artefacts are there, in plain sight.
I have a commissioned shoot coming up in July of a large mixed-use apartment building. Typically, I obtain a city permit to barricade the parking directly in front of the building. A tall commercial truck or van in front can really diminish important views. Do you typically go this route or do you now depend upon AI tools to remove vehicles? My concern in this particular shoot is that because I will be so close to the building, removing large vehicles can be very time-consuming and difficult. Removal errors become quite obvious this close.
Thanks for your note, Andrew. This is a real problem for all of us. And AI is not always going to be the answer, especially if the truck/van takes up significant areas of the building.
Here is what I have done in the past. If it’s a private building, I would work with the firm that contracted the photography (architect, contractor, building management company….whoever) and, as you have done, have them arrange to cordon off the critical parking/driveway areas for the time that you expect to be photographing. It’s in their best interest to do this. It is easier to do this with a morning shot. If it’s a public area it's clearly harder to arrange. And it may force you to photograph at a less-than-optimum time. Also check to see if a weekend session precludes the vehicles being there.
Absent any help, I have blocked driveways, parking areas, etc myself with yellow tape and have had my assistants there to direct traffic away from the building. People complain; let them.
Here is a last resort, which I have also used. It helps if it’s an axial shot. I take the photograph with the vehicle in the way. If possible, I then go between the vehicle and the building and shoot an additional view of what is being blocked at exactly the same height as the overall photo, with the same lens and the same exposure settings. Then, I very carefully layer the two images. If the light conditions are identical (e.g., the van/truck is not casting a shadow) this can actually work. The result will probably be better than trying to use AI to recreate what’s behind the van.
@@stevenbrookephotography Great thoughts, Steven. One downside to barricading parking is people ignore the signs!
Your last idea is interesting. I haven't tried that approach for subjects of this nature. Will have to consider that.
LOL They usually don't plow through yellow tape. And having an assistant can help. Here is a tip: if you can get to them before they turn off their engines, you can usually get them to park elsewhere. Once the engine is off, the inertia is apparently too much for them. Good luck with this, Andrew. We are rooting for you!
So as a professional photographer how do you feel about relinquishing your property rights to adobe when you agreed to the new TOS? Which also includes blanket access and domain to all the videos, and any other work that you have used an adobe product on.
I would say that zooming up on closer inspection these tools leave a lot to be desired - strange artifacts and often unconvincing removal - looking at images not zoomed in might look ok but zoom in 100% and things often fall apart
Yes. Close inspection of the final result is certainly necessary. In many cases, it takes more than one pass with either tool to achieve a totally believable result. Generative Fill, in particular, can often create it's own structures and requires careful inspection following the manipulation. That said, I believe that these tools will continue to evolve and that the results will correspondingly improve.
I think sometimes you need to show places warts and all so as not to misrepresent. Holiday brochures etc.
I agree. In my work documenting Miami's endangered historic neighborhoods, I make a point of not removing anything. All the encroaching high rises, out-of-scale development, and other urban detritus are part of the story. But this project is more photojournalism than work for architects or designers, documenting new residences or commercial buildings.
I'm done with Adobe Stuff. I rather learn other tools and learn to complement the AI with stable diffusion. Painful Process but must be done. In the meantime So far I've have not find something that I could not do with other tools like GIMP or Davinci Resolve. (Yes, you can Edit pictures, not just edit video). Resolve Studio has a tool for wire removal and other for object removal that does exactly the same. Meaning is content aware.