Dear mathphotographer, Thank you for these wonderful videos. I enjoy your beautiful content and your delivery is just as amazing. I always learn something new and valuable here. You are AWESOME!
The real issue, for me, about twenty four megapixels is to do with how we actually see. As we get older our eye sight goes from a hundred megapixels to twenty four but at the same time, our perception, our sense of 'what we see' goes up by much more. Painters, the ones we come to love, did not concern themselves with such matters. Instead, they sought to create images that pleased our eyes when viewed from a similar distance to reality. That is to say, from perhaps three meters onwards. Twenty four megapixels, in my humble opinion, approximates what we can see but more importantly, how we perceive reality which is, ultimately, impressionistic. The sharper and image, the more intense the detail, the less we tend to 'see'. A paradox.
Ultimately , the ability of the software to upscale the file in order to get that great result has got to be tracked down to the amazing performance of the SL lens. What a beast.
As usual an excellent video. I have Gigapixel software & use the SL2 multiple shot mode, one thing you didn't point out firstly the noise suppression can be increased in the software & the software also added some sharpening very noticeable. For best results I agree that multishot is generally better but limitations of having to have a tripod & time, but generally if I use the Topaz software I use the added packages Topaz Sharpne & DeNoise when you use these very difficult to tell the difference. Topaz have released a new package call Photo AI which combines all three packages & is almost as good as the septette packages. I now use this as a standard on all my Leica images from the Q@M, SL2 & M11 with much better results as the finished product.
Thank you for the video. I have been using Topaz GigaPixel AI and also Topaz Video AI since the start and the improvements have been tremendous since the apps were first released. Give it a few more years and camera resolution will literally not matter anymore with software like this that can re-scale the image to whatever you want as the AI engines get more and more sophisticated. I shoot a lot with my 16MP Leica X-Vario that when upscaled 2x to 64MP, produces files similar to my 61MP Sony a7RV. This is crazy stuff and will only get better! We are starting to witness in real life what in Ridley Scott's 1982 Blade Runner was just science fiction during the scene with Harrison Ford enhancing the victim skins images. Amazing stuff..
The great advantage of Gigapixel, of course, is that you need not concern yourself with motion in the image the way you must with pixel shift. I find that Gigapixel works best coupled with sensors sporting high quality pixels. Some of the best results I have seen from this software has come when I used it on files from my Nikon D3S. That camera "only" had 12MP, but they were very high quality pixels with lovely tonal rendition. Gigapixel produces beautiful results when up-sampling those 12MP files to roughly 45MP. The images are about as good as any current native 45MP FF camera. Remember, too, that Gigapixel is improving every year, so expect your final results to get better and better as future iterations are released.
Very interesting video! Topaz has been working intensely on developing their AI application, which combines denoise, gigapixel and sharpen functionality. Every week they provide an update. It would do possibly do as good a job as Gigapixel to enlarge your image, as well as control the noise. I use it a lot. Incidentally, I added the SL2-S to my collection due to your recommendation and I love it with my M lenses.
Many thanks, glad you enjoy your SL2-S, its a great camera with astonishing image quality and, nevertheless, so easy to handle with a well structured menu and access to all SL and M lenses.
I’m wondering if using one of these two techniques this would allow you to make larger prints with the sl2-s? If so, how large do you think? Impressive video, thanks.
Thanks for another very interesting and informative video. I was wondering how you filmed the panning motion at 8:48 in the video. It almost looks like it was a drone shot.
Of course doable and I experimented with the software a lot, Edward. Processing time when the AI models update is longer if you go for higher resolution images as source file, of course.
I wonder if the 96MP Gigapixel AI results will be the same if you convert from a 24MP original (like you did in this video) vs. either a 12MP or a 47MP original. As for this video, I thought the GP AI image looked a tiny bit sharper than the multishot image.
That looks like a great result. Almost matching the multishot mode, which is one of my favorite options in the SL2 and S1-cameras. Thanks for the comparison. Thinking about longer exposure times, do you know whether the SL2 IBIS works together with the OIS in a Lumix lens such as the 24-105? I’ve somewhere read it does, but could not get confirmation from Leica… Thanks for your thoughts on this.
@@mathphotographer I’ve tried this over the past week. A Lumix 24-105 on my SL2-S. As far as I can judge the OIS and IBIS work well together. I managed a sharp handheld shot at 1s on approx. 35mm. In the Camera Settings menu the SL2-S recognizes the lens profil. Based on the OIS switch on the lens, you can set the IBIS. When OIS is switched on, the IBIS is grayed-out in the ON position (which makes sense). When you switch off OIS, you can either leave IBIS on or switch it off as well. I personally like the Lumix S 24-105. It combines a wide range, quite good IQ and great close-focus capabilities. A very universal lens which only costed me €850 brand new in bulk package. Great to have as a walk-around or travel lens, next to a couple of (APO) primes. There’s no real Leica or Sigma alternative for it. The 24-90 is almost double the weight and 4 times the price.
@@jacobh5817 Thanks Jacob, as expected. Lumix lenses belong to the L-Mount Alliance so I expected everything to work fine as you confirmed, thanks for that. And you are of course spot-on right, price vs performance of the Lumix zoom lens is outstanding!
Not only the bushes, all the paraphernalia (like antennae) on the rooftops looks better in the multishot image. Gigapixel AI seems to be good at replicating surfaces, but less so at enlarging fine structures.
Thanks Barry. The Nikon Z9 does not have Multishot but other cameras have, so a comparison of the Leica Multishot with Sony or Fuji Multishot is doable. In addition, a comparison between the 24MP Leica SL2-S image as Multishot versus the 45MP Nikon Z9 image is also doable.
Maybe it's me but I am seeing artifacts and a painterly effect in the Gigapixel shot I don't like - however, that is pixel peeping so maybe up to certain print sizes you won't really see it (but of course you use Gigapixel when you want to print large, so.....)
Depends indeed on the print size whether you notice or not. What I found astonishing is that GigaPixel AI generally delivers better results than a simple upscaling by Photoshop, really an interesting software and fun to play with.
Thank you for the video, I wonder for printing portrait (large print) how would the skin colour look, because I noticed when using high pixel camera, the skin colour for people like me (Arab colour skin) looks nicer than when I I use 24 megapixel camera. Would the software do the same job for the skin colour
Hi Bassam, not sure since I tried GigaPixel AI so far only for landscape and architecture. I will play a bit with people images and then see what the results are.
Another possible factor that comes to mind, given Multi-shot's sensitivity to movement, is that even with the camera on a tripod or solid piece of furniture, such a high vantage point might result in some slight movement of the building affecting the quality of the final image.
Correct Paul, Multishot is definitely superior in a calm still scene. But if you have moving subjects, you get some workaround with GigaPixel AI which is not as good as Multishot but Multishot would have to deal with artefacts and motions and GigaPixel AI will just scale up the image without getting you into troube with motions.
I often photograph birds with my Hasselblad X2D 100C using the XCD 135 + 1.7x converter. Sometimes the birds are just a little too far away to produce a decent image so I tried out the Topaz GigaPixel (in demo mode with the watermark) to see if it could enhance some of these images. Quite frankly I found it to be absolute rubbish. The resulting images were so full of artifacts and noise that it was better to just upscale the image in my old Photoshop 7.1. Certainly not impressed. Ok, I may not have given the software a thorough testing and tweaked the many settings but I don't have the time for that. Btw - I love my X2D.
The X2D is one of my favorites :) I can imagine that for bird photography an enlargement by Software is not giving you the desired outcome. For cityscapes its different as you saw, easier to scale up than a tiny bird at the horizon. Thanks for sharing, Roger.
@@mathphotographer Thank you for your reply; you are very kind. I agree with you that this software can be useful as you demonstrated. However for me a new software innovation has to work out of the box for me to bother with it. It was just an idea that I thought might work. Now a piece of software that did work "out of the box" was Helicon Focus which although I don't use very often works quickly and efficiently and produces stunning "3D" photographs. Kind regards.
Indeed Gigapixel AI creates details that don't exist, when AI recognizes the subject part to blow up. I have an example with far away trees with leaves that have only several pixels in size with no details on the native image. Gigapixel recognizes these parts as leaves and adds leaves structure details to the blown up image. I guess a building concrete wall in the night is not part of the AI models bank and so it is not able to add concrete structure. Gigapixel continuously improves and implements the AI models bank and with time the software becomes better and better. As Gigapixel is really adding non existing subject details it's prohibited by cultural legacy institutions that only rely today on 400 Mpix H6 100d multishot (or Phase One) gear
Absolutely not, of course not. Multishot that is done by combining multiple native resolution photos pixel moved by the shifting the sensor, has nothing to do with software upscaling from a single shot and that requires extremely powerfully computing power to render that no camera would ever be able to do.. at least not yet at this point.
Hi Maurizio, the Leica SL2-S has a "true" Multishot and not upscaling like GigaPixel AI. The SL2-S uses sensor shift to create up to 8 consecutive frames and the information of each of these frames contributes to the overall 96MP image. So no marketing, true technology :)
I only ‘asked’ because, just like you said at the beginning of the video, GigaPixel could only upscale the information present in the source file. Comparing the upscaled image with the multishot image revealed it was 99% the same! Hence my suggestion… Anyway, this comparison shows how good GigaPixel really is, so well done.
The New York image is fantastic. Thanks for another interesting video
Pleasure :) The NY image is on the wall in our dining room and always makes me believe I have a window into NYC ...
Dear mathphotographer, Thank you for these wonderful videos. I enjoy your beautiful content and your delivery is just as amazing. I always learn something new and valuable here. You are AWESOME!
Many thanks, Bah Aram, your positive feedback is much appreciated.
The real issue, for me, about twenty four megapixels is to do with how we actually see. As we get older our eye sight goes from a hundred megapixels to twenty four but at the same time, our perception, our sense of 'what we see' goes up by much more. Painters, the ones we come to love, did not concern themselves with such matters. Instead, they sought to create images that pleased our eyes when viewed from a similar distance to reality. That is to say, from perhaps three meters onwards. Twenty four megapixels, in my humble opinion, approximates what we can see but more importantly, how we perceive reality which is, ultimately, impressionistic. The sharper and image, the more intense the detail, the less we tend to 'see'. A paradox.
Ultimately , the ability of the software to upscale the file in order to get that great result has got to be tracked down to the amazing performance of the SL lens. What a beast.
Agree, thanks!
As usual an excellent video. I have Gigapixel software & use the SL2 multiple shot mode, one thing you didn't point out firstly the noise suppression can be increased in the software & the software also added some sharpening very noticeable. For best results I agree that multishot is generally better but limitations of having to have a tripod & time, but generally if I use the Topaz software I use the added packages Topaz Sharpne & DeNoise when you use these very difficult to tell the difference.
Topaz have released a new package call Photo AI which combines all three packages & is almost as good as the septette packages. I now use this as a standard on all my Leica images from the Q@M, SL2 & M11 with much better results as the finished product.
Thanks for sharing, Peter. It is an astonishingly well working software package!
Thank you for the video. I have been using Topaz GigaPixel AI and also Topaz Video AI since the start and the improvements have been tremendous since the apps were first released. Give it a few more years and camera resolution will literally not matter anymore with software like this that can re-scale the image to whatever you want as the AI engines get more and more sophisticated. I shoot a lot with my 16MP Leica X-Vario that when upscaled 2x to 64MP, produces files similar to my 61MP Sony a7RV. This is crazy stuff and will only get better! We are starting to witness in real life what in Ridley Scott's 1982 Blade Runner was just science fiction during the scene with Harrison Ford enhancing the victim skins images. Amazing stuff..
Or like in "Blow Up" from 1966.
@@bmwohl Yep, that's right!☺☺☺
Thanks Android, your comment is spot-on, its remarkable what AI can do today and of course it will get better over time. Technology is fascinating!
The great advantage of Gigapixel, of course, is that you need not concern yourself with motion in the image the way you must with pixel shift.
I find that Gigapixel works best coupled with sensors sporting high quality pixels. Some of the best results I have seen from this software has come when I used it on files from my Nikon D3S. That camera "only" had 12MP, but they were very high quality pixels with lovely tonal rendition. Gigapixel produces beautiful results when up-sampling those 12MP files to roughly 45MP. The images are about as good as any current native 45MP FF camera. Remember, too, that Gigapixel is improving every year, so expect your final results to get better and better as future iterations are released.
Agree, thanks for your comment, Robert.
Very interesting video! Topaz has been working intensely on developing their AI application, which combines denoise, gigapixel and sharpen functionality. Every week they provide an update. It would do possibly do as good a job as Gigapixel to enlarge your image, as well as control the noise. I use it a lot. Incidentally, I added the SL2-S to my collection due to your recommendation and I love it with my M lenses.
Many thanks, glad you enjoy your SL2-S, its a great camera with astonishing image quality and, nevertheless, so easy to handle with a well structured menu and access to all SL and M lenses.
One more great video. A very
Clear lesson.
Thank you Duc.
would be cool to see a real 96mp camera vs the 24mp on GigaAI to really see what it can do in real pixel count. Thanks for sharing!
I’m wondering if using one of these two techniques this would allow you to make larger prints with the sl2-s? If so, how large do you think? Impressive video, thanks.
Great topic! Thanks for the comparison.
Glad you liked it.
Great video, many thanks
Interestingly I was considering the purchase of the Reporter today, maybe your video is a special sign 😊
Thanks Ronald - maybe indeed a good sign for your Reporter acquisition :)
Outstanding video!
Thank you very much!
Output images look very nice on both.... 'Gigapixel' is reminiscent of 'Fractal Images' software over 20 years ago.
Good comment and observation, thanks Stuart.
Cool idea, need to try that as well with mine (and I had some Benro Polaris lying around which I also bought for the Gigapixel/pano idea)
thanks for the videos Maybe now I can decide which system to buy.
Glad I could help.
Thanks for another very interesting and informative video. I was wondering how you filmed the panning motion at 8:48 in the video. It almost looks like it was a drone shot.
How about scaling up a SL2 47mp image vs 100mp from GFX100s or X2D?
Of course doable and I experimented with the software a lot, Edward. Processing time when the AI models update is longer if you go for higher resolution images as source file, of course.
Always appreciate the simplicity of Leica interface when I see it in action. Does GigaPixel AI have option to work with RAW files?
Yes I tried it out, Donald. You can import RAW images and work with the better information than what you get with a JPG.
I wonder if the 96MP Gigapixel AI results will be the same if you convert from a 24MP original (like you did in this video) vs. either a 12MP or a 47MP original. As for this video, I thought the GP AI image looked a tiny bit sharper than the multishot image.
You could feed the 24Mpx DNG into Topaz AI and Lightroom also has an upscaling / enhance function built in , so you don't even need Topaz.
It’s different upscaling. GigaPixel AI uses enhanced AI methodology which Lightroom has not built in
That looks like a great result. Almost matching the multishot mode, which is one of my favorite options in the SL2 and S1-cameras. Thanks for the comparison. Thinking about longer exposure times, do you know whether the SL2 IBIS works together with the OIS in a Lumix lens such as the 24-105? I’ve somewhere read it does, but could not get confirmation from Leica… Thanks for your thoughts on this.
Thanks Jacob. I think the Lumix lenses cooperate with the SL cameras but will check next days.
@@mathphotographer thanks! Much appreciated.
@@mathphotographer I’ve tried this over the past week. A Lumix 24-105 on my SL2-S. As far as I can judge the OIS and IBIS work well together. I managed a sharp handheld shot at 1s on approx. 35mm. In the Camera Settings menu the SL2-S recognizes the lens profil. Based on the OIS switch on the lens, you can set the IBIS. When OIS is switched on, the IBIS is grayed-out in the ON position (which makes sense). When you switch off OIS, you can either leave IBIS on or switch it off as well. I personally like the Lumix S 24-105. It combines a wide range, quite good IQ and great close-focus capabilities. A very universal lens which only costed me €850 brand new in bulk package. Great to have as a walk-around or travel lens, next to a couple of (APO) primes. There’s no real Leica or Sigma alternative for it. The 24-90 is almost double the weight and 4 times the price.
@@jacobh5817 Thanks Jacob, as expected. Lumix lenses belong to the L-Mount Alliance so I expected everything to work fine as you confirmed, thanks for that. And you are of course spot-on right, price vs performance of the Lumix zoom lens is outstanding!
Thnx! 👍🏼✨
Pleasure :)
Not only the bushes, all the paraphernalia (like antennae) on the rooftops looks better in the multishot image. Gigapixel AI seems to be good at replicating surfaces, but less so at enlarging fine structures.
Agree, the Multishot is better as said in the video. Nevertheless astonishing what Software can deliver today. Thanks for your comment.
I found the MultiShot most impressive. Could you compare MultiShot on the Leica with something similar on the Z9?
Thanks Barry. The Nikon Z9 does not have Multishot but other cameras have, so a comparison of the Leica Multishot with Sony or Fuji Multishot is doable. In addition, a comparison between the 24MP Leica SL2-S image as Multishot versus the 45MP Nikon Z9 image is also doable.
Maybe it's me but I am seeing artifacts and a painterly effect in the Gigapixel shot I don't like - however, that is pixel peeping so maybe up to certain print sizes you won't really see it (but of course you use Gigapixel when you want to print large, so.....)
I agree. The architectural elements all appear to have a Photoshop "glowing edges' filter applied.
Depends indeed on the print size whether you notice or not. What I found astonishing is that GigaPixel AI generally delivers better results than a simple upscaling by Photoshop, really an interesting software and fun to play with.
Thank you for the video, I wonder for printing portrait (large print) how would the skin colour look, because I noticed when using high pixel camera, the skin colour for people like me (Arab colour skin) looks nicer than when I I use 24 megapixel camera.
Would the software do the same job for the skin colour
Hi Bassam, not sure since I tried GigaPixel AI so far only for landscape and architecture. I will play a bit with people images and then see what the results are.
Multi-shot was definitely superior. Could have used a pinch more sharpening to my eye but that could be CZcams compression.
Another possible factor that comes to mind, given Multi-shot's sensitivity to movement, is that even with the camera on a tripod or solid piece of furniture, such a high vantage point might result in some slight movement of the building affecting the quality of the final image.
Correct Paul, Multishot is definitely superior in a calm still scene. But if you have moving subjects, you get some workaround with GigaPixel AI which is not as good as Multishot but Multishot would have to deal with artefacts and motions and GigaPixel AI will just scale up the image without getting you into troube with motions.
I often photograph birds with my Hasselblad X2D 100C using the XCD 135 + 1.7x converter. Sometimes the birds are just a little too far away to produce a decent image so I tried out the Topaz GigaPixel (in demo mode with the watermark) to see if it could enhance some of these images. Quite frankly I found it to be absolute rubbish. The resulting images were so full of artifacts and noise that it was better to just upscale the image in my old Photoshop 7.1. Certainly not impressed. Ok, I may not have given the software a thorough testing and tweaked the many settings but I don't have the time for that. Btw - I love my X2D.
The X2D is one of my favorites :) I can imagine that for bird photography an enlargement by Software is not giving you the desired outcome. For cityscapes its different as you saw, easier to scale up than a tiny bird at the horizon. Thanks for sharing, Roger.
@@mathphotographer Thank you for your reply; you are very kind. I agree with you that this software can be useful as you demonstrated. However for me a new software innovation has to work out of the box for me to bother with it. It was just an idea that I thought might work. Now a piece of software that did work "out of the box" was Helicon Focus which although I don't use very often works quickly and efficiently and produces stunning "3D" photographs. Kind regards.
Indeed Gigapixel AI creates details that don't exist, when AI recognizes the subject part to blow up.
I have an example with far away trees with leaves that have only several pixels in size with no details on the native image. Gigapixel recognizes these parts as leaves and adds leaves structure details to the blown up image.
I guess a building concrete wall in the night is not part of the AI models bank and so it is not able to add concrete structure. Gigapixel continuously improves and implements the AI models bank and with time the software becomes better and better.
As Gigapixel is really adding non existing subject details it's prohibited by cultural legacy institutions that only rely today on 400 Mpix H6 100d multishot (or Phase One) gear
Thanks for sharing, Jean-Claude, interesting comment.
Could it be possible the SL2-S simply has Gigixel inside and this whole “Multishot” is marketing BS?
Absolutely not, of course not. Multishot that is done by combining multiple native resolution photos pixel moved by the shifting the sensor, has nothing to do with software upscaling from a single shot and that requires extremely powerfully computing power to render that no camera would ever be able to do.. at least not yet at this point.
i think you are BS.
Hi Maurizio, the Leica SL2-S has a "true" Multishot and not upscaling like GigaPixel AI. The SL2-S uses sensor shift to create up to 8 consecutive frames and the information of each of these frames contributes to the overall 96MP image. So no marketing, true technology :)
I only ‘asked’ because, just like you said at the beginning of the video, GigaPixel could only upscale the information present in the source file. Comparing the upscaled image with the multishot image revealed it was 99% the same! Hence my suggestion… Anyway, this comparison shows how good GigaPixel really is, so well done.