Peter, thank you for this, it's a good topic. But IMO a modest difference in sharpness is adjustable just like contrast. What really matters is ultimate resolution/detail, noise, and dynamic range. Look for resolution limits at the tiniest detail levels and compare which one starts to fail or breakdown first, right down to pixel level. (Include the standard photo image as well for this; just resize to 50 and 80 using standard Lanczos resize in most image editors). Then, to compare dynamic range and noise, boost the deepest shadows by five stops in post, and look at how they retain detail and how much noise (also compare standard photo). You could repeat the test at higher ISO to see if there are more differences. It is worth remembering that AI/ML is making things up to fill in the gaps, and sometimes it will get it perfect and sometimes it will put rubbish in there. We almost have to look at every piece of detail over the entire image to see which it has got right and which bits it has got wrong. Cheers
One of the uses for Gigapixel is increasing the size of an image after a significant crop. Another is for really small images you "steal" off the internet or from video stills.
One of the more interesting photo comparisons Peter. I have used Gigapixel since its first release. You have just confirmed my thoughts that I do not need to upgrade my old E-M1 to get pixel shift. One thing I will say though is that on hot days when there is air turbulence gigapixel can give strange results with up resing of far off buildings and trees in localised heat shimmer on single shot panoramas (have not tested multi shot panos). All confirmed with pixel peeping the original shot, but not immediately apparent when viewed normally.
Good review. I used the trial Gigapixel on my E-M5 MkII images (and others) as well and found the quality of upsizing was surpringly good and the files had no problems with further processing. Was also fun taking my 3MP shots from 10 years ago and up sizing to 12 MP and they looked good as well!
Thanks Peter. A very interesting comparison. I recently tried Topaz Sharpen AI and that worked very well. Really enjoyed your stream with Rob the other day. Hope you do another one soon 👍
I use both, I actually use Gigapixel to down size and sharping my HHHR shot, I also use denoise AI, the HHHR shot have many other advantages , such less noise, more DR, and longer exposure time.
Hi Peter, great video as usual, just a heads up, The Olympus Pen F also has the high res shot feature, Images shot at 50mp produces the same 80mp file.
I got gigapixel a few months ago and I’m amazed by how well it upscales high quality files. I’ve created 90+ mp files out of iso 100 5D mkiii files that have amazing detail and no weird artidacts as far as I can tell. Realistically, creating approximately 50mp files from Something like R6 mkii that have as much detail and sharpness as R5 is feasible.
Great informative comparison Peter. I've been using all the Topaz Software for decades. I'm seriously considering the Olympus because of the Live Comp mode
I've been using Topaz Denoise AI as an "additional external editor" with Lightroom along with Sharpen AI. Denoise seems often to be simpler and better. Sharpen does offer rather contrasting options. In terms of image quality, I just had an old Olympus four-thirds 35-100 mm f/2.0 lens out in use.
High Peter. Nice comparison and surprising result as a high res shot does include more information than a single shot. Another thing: you forgot to mention than PEN-F does also make high res tripod (not handheld).
@@ForsgardPeter After watching your video, I find that the High-Res function is probably best as a "scanner" or for product photos. For landscapes, Topaz Gigapixel is probably better, and then it fits with my idea of buying a Fuji GFX 50R. No need for 100S. The GFX 50R fits perfectly with the m43 cameras I already have, even the optics can both use, together with Metabone's adapters.
@@ForsgardPeter That's a pity, since that's a very nice camera which you can now get quite affordable 2nd hand. (Although I prefer the usability of OM-D and the highres shot from the Pen-F cannot deal with movement such as in water surfaces. That seems to be improved in later Olympus models.)
I was waiting for Topaz Labs discounts and then bought one, then another of their software. Eventually, I bought enough that at another discount where they offered a Suite discount, I bought the rest of the modules. The Gigapixel AI and the Denoise AI are very good for someone who does not have intense Photoshop skills or does not have the time/effort to carry out high skilled Photoshop equivalents. The enhancements do help lift some of my modern images but I was really happy with Topaz Labs software for enhancing 3 MP images from my first digital cameras, the Nikon Coolpix 775 and the later from my Olympus C-750uz and Kodak P880.
I have many of the Topaz products including AI Gigapixel. They have great products but you have to be careful with them and take a good look at the output.The latest version of Gigapixel AI can be used as a plugin for Lightroom and Photoshop. I haven't tried it yet, I assume it works. Apparently you did not do an upscale conversion but a 1:1? I have never tried that, I always use it to enhance very cropped wildlife images. Sometimes a 6:1 upscale. I get mixed results. Sometimes amazing other times obvious image issues. I haven't used the latest version yet and they do make improvements all the time. I would add that the DeNoise AI is really good for hi ISO shots with lots of noise. It does an amazing job. Anyway, thanks for your videos. Always interesting to watch.
Thanks, Peter. I've tried Gigapixel some time ago with good results, but ended up buying Denoise AI and Sharpen AI. My M10 mark II can now produce much cleaner and sharper images. Even my E-3 now is worthy again! I think this technology is capable of blurring the difference between high end and enthusiast /amateur equipment. It may be interesting to compare, for example, a picture taken with an M1 with pro lens against a picture from an M10 with kit lens processed with Denoise and Sharpen.
Thanks Peter! The conclusions of your research is astonishing. What would be also interesting. DPReview use test image for all cameras they test. In the right lower corner of the image there are tubes of painting colors with letters and the letters are not discernible even in some HiRes images (if available in particular camera). It's interesting would Gigapixel applied to normal image beat performance of HiRes image judging upon the letters on the tubes? In general, test using text, paper for instance, would be more consistent than city landscape because in case of the paper text there is a strict differentiation - you either able to read the text or not. The paper has to be placed on the distance from camera such that in normal image the text is not completely recognizable or barely recognizable but not too long distance when the size of letters in pixels is too small - not less than 6 pixels height - to avoid discretization impact.
Ah, the life of a professional photographer these days... can't even afford a pair of jeans with both knees intact....... : ) This was a very interesting video.... since I only have an EM-1ii, I think I will invest in Gigapixel.... I plan to print some of my favorite landscape pictures so this will help keep them looking good. Thanks for sharing Peter!!
Look out for heat shimmer on landscapes. It may be unnoticed in the original but Gigapixel will show it up in strange ways such as a bent wall or window or a small cluster of warped trees. Having said that, its great on landscapes.
Hi Peter Nice review and surprising result indeed. But what would be funny to check is how a high res shot could be upscaled and sharpened to let's say to 125 (handheld) or 200 (tripod) MP, because mathematically it's an evidence that a high res shot contains more information than a single shot, as it's a merge of 16 (random move) or 8 images (controled move), depending handheld or tripod. We have to remember that any camera can produce hand held high res shot (even the PEN-F 😉), if you take a burst of RAW images and merge them in post with photoshop (set minimum setting for any NR/sharpenig in RAW, open as calques, align, convert to dynamic object with median rendering, sharpen - it works well with Topaz Denoise AI which can also sharpen).
Hi Peter, thanks for sharing this interesting tests. Is it possible to use gigapixel AI on HR files to get even better results? Not sure if it's possible. Thanks again.
Peter, since Finnish people are famous for being great hockey players, you should go to a rink and take some action shots of them skating on the ice. It would be a great way to see how Olympus cameras perform in sports situations.
Thanks for the comparison. Looks like the ultimate in image quality could use a combination of the two, blending in the Gigapixel result of moving objects with the stationary clarity of in-camera processing.
Whau! Right on time :) I was looking on Topaz software and was wondering how good is they Gigapixel? I were planing to do my own comparison with HiRes and here you are :) with your video! And looks like is definitely worth to consider. Peter, thanks a lot for video!
Peter, after taking a High Rez shot with Olympus, did you take it a step further and process it in Topaz? Does the High Rez file from Olympus stand up for further resolution up scaling in Topaz? It would be great to know what the result was. I would like to continue with my Olympus lenses and 20mpx bodies rather than make an expensive gear change to Full Frame 40 or 45 mpx bodies.
I have used Gigapixel to enhanced old snapshots taken from 2009 video and it works quite well albeit a little slow (3 minutes per photo). I then finished them in Lightroom.
My prediction: the trend of cameras is not bigger sensors, but smarter image processing technologies (e.g. AI); bigger sensors bring us less and less advantage. Smart cameras will emerge. This can be a promising future of MFT system. Some image processing software packages will be user installable, just as we install apps in smart phones. Once you install a package (e.g. something like Gigapixel AI), your camera gains a special ability. Of course, for the functions which need mechanical operations, such as driving the lens to zoom in/out, or to change the aperture, or to change the focus, we may still need firmware upgrade to support it, unless APIs are exposed to external programmers. But more and more visual effects can be achieved using pure computational approach, instead of optical approach, so less customized mechanical operations will be needed. Less firmware, more software.
AI is high energy consumption thing and this is main obstacle to put it inside camera. Also AI adoption in camera contradicts trend on megapixel increase.
@@sstansm7f True. But phone cameras have already benefited a lot from AI (e.g. portrait mode, night mode, etc.) and the small sizes of lenses. If MFT does not go the same direction, it won't be able to survive between full frame and phone cameras. Regarding the energy consumption, the battery life of mirrorless cameras is shorter than DSLRs, but mirrorless still wins the market.
@@ruiduan2843 Yes, but 1) Smartphones produce low quality initial images (apprx as P&S digital compact cameras in 2004-2010) it does not require much CPU resources to enhance them 2) Smartphones enhance image on the fly as one can see them on display before taking picture 3) Smartphones does not allow serial shooting with measurements. One can add AI capabilities to cameras as post-processing via menu but the problem is that the photographer can't assess resulting image unless camera will have FHD display 6" size.
Interesting comparison Peter. I find software such as DeNoise invaluable in high-contrast situations such as sunrise and sunset where HDR isn't practical and the shadows need to be pushed in post, resulting in noise. That said, from a creative perspective I think we need to draw the line at excessive digital manipulation e.g. sky replacement or combining elements from different locations i.e. how do we define a photograph vs a digital art.
That's easy. A photograph is not manipulated. If you do add or take away objects in a photo, it's no longer a photo, but something else. Nothing wrong with that, if that's what you like. But I don't like people who pretend that a composited image is a photograph. It isn't.
Not very often if I am going to photograph. For video work, I need one. My favorite tripod at the moment it is the Peak Design Travel Tripod. I have not made any videos about tripods.
There are so many options and how they compare varies from photo to photo and area to area within a photo. Check this video from 17:09 to 18:30. th-cam.com/video/5LrPKoNvPPc/w-d-xo.html . You never know what error is just around the corner with these machine learning upsizing apps. That was my original point: they are inventing more information than they are keeping. Unlike the Olympus technologies that are compositing real data captured in the field from the subject. Also, if you want the ultimate in high resolution, panning is the way to go. Machine learning upsizing is a toy compare to that. If you want to play with the toys, something vastly cheaper and built into a powerful editing app like Pixelmator makes a lot more sense to me.
Thanks for doing this comparison, Peter. I own Gigapixel AI, but never bothered to compare it against my E-M5 MkIII handheld hi-res. The good thing is that Topaz is constantly tweaking the AI algorithms, so its output will keep getting better. I see that Skylum has gone all-in on AI with their upcoming "Luminar AI" product. (Question: In your tests, couldn't you simply have dialed in "2.5X" and "4X" in the "Scale" setting of Gigapixel to get your 50MP and 80MP output?)
Another advantage of Oly hi-res modes is noise reduction through averaging. Now if you really want a lot of pixels, create a tripod hi-res panorama then double it with Gigapixel AI. Btw the current version of gigapixel ai does work from Lightroom, at least on my Mac. I think Anthony Morganti did a video on that.
@@ForsgardPeter MFT drew me in. I had been saying that I would skip 'full frame' and go to medium. Results from MFT are better than I dreamed. I imagine 'full frame' is too. I don't plan on big prints, but often crop minimalist details out other wise usable prints. These areas can fall victim to frindging, which I have been inept at dealing with. With my first PC in 25 years and only Windows 10 home , the search is on for software. I do not have internet connection, so no subscriptions. Thanks Chris
The poor corner sharpness was due to the lens optical limitation. You need really good glass that is shot at optimal aperture to take full advantage of high-res shots.
You should, Jimmy! Their DeNoise AI and Sharpen AI products in particular are very good. Sharpen AI can even rescue slightly out of focus shots! And it's amazing in correcting for handheld shake.
No because of the remove blur slider it produces smooth images. If you work with a real small starting image, it will look painterly, too smooth but better than nothing.
A very strange comparison, indeed! By the way Gigapixel Ai does a great job on old photos, not just right now photos! And eventually Ai will be the downfall of our style of editing.
Good review.But IMHO more correct is comparing for example Lightroom interpolation tool vs Topaz Ai,while both do "math" and fighting resizing artifacts ,but actually not add any new details.Olympus multi-shot mode providing true 50/80mpx and no brainer it will be better than Topaz Ai.Again IMHO.
It's amazing what image ai can do nowadays. But how often does the average photographer a high res picture? For $100,- it's expensive if you use it as a gadget.
It's less than $100, but it is more expensive than most of their other AI-powered apps. It's real strength is upscaling a highly cropped image back to original size, or squeezing more details from scanned photos or slides. You would rarely use it to upscale a typical 20MP photo, because there's no point (and the TIFF files created are huge!).
Hallo Peter! I made some tests. If I shot with my EM-1 III in Handheld HighRes mode and do sharpening mit Topaz DeNoise AI (Yes in AICear Mode) the results are much, much better as in Topaz Gigapixel AI. Much more detail and more sharpness.
The big fear, and some predict (Steven Hawking, and a lot of others), that AI will ruin us and the planet. We'll end up in a battle for resources. We should nip it in the bud.
@@sstansm7f Look at Luminar 4 or AI. You can easily do edit portrait with some sliders. Most clients wouldn't pay you for more effort when doing this with Photoshop. That sounds like a benefit. The problem start, if all clients can do this and all the magic happens on a smartphone. By now, you mostly know which pictures are composed and what are real, especially from non photographers. But if your smartphone replace the sky, eyes, moon, sun and simulate the lightning to the background, you have to keep up against that AI composed colorful noise free pictures. The good think is, the AI could not guide a group of persons, could not shoot a concert or wedding and is not able to help you for creative work.
@@polarized8708 Smartphones take standard pictures and capable to enhance them to the level of acceptance for average customer. It's mass product. For average people. Also smartphone made pictures are too noised and colors degrade in low light conditions far behind what cameras can produce.
I’ll have to take your word for it, because I cannot see any difference between the images on this video. I bet if you adjusted the contrast and perhaps sharpened the Gigapixel AI image ever so slightly, it would be an even closer match.. there’s no way anyone would be able to tell.
Peter, the 80 mpix high res raw files are always very soft and need to be downsized and sharpened to look acceptable IMO. I assume you did not downsize the 80 mpix file, however was it sharpened and if so what settings and in which program.
Get Gigapixel AI: topazlabs.com/gigapixel-ai/ref/644/
Peter, thank you for this, it's a good topic. But IMO a modest difference in sharpness is adjustable just like contrast. What really matters is ultimate resolution/detail, noise, and dynamic range. Look for resolution limits at the tiniest detail levels and compare which one starts to fail or breakdown first, right down to pixel level. (Include the standard photo image as well for this; just resize to 50 and 80 using standard Lanczos resize in most image editors). Then, to compare dynamic range and noise, boost the deepest shadows by five stops in post, and look at how they retain detail and how much noise (also compare standard photo). You could repeat the test at higher ISO to see if there are more differences. It is worth remembering that AI/ML is making things up to fill in the gaps, and sometimes it will get it perfect and sometimes it will put rubbish in there. We almost have to look at every piece of detail over the entire image to see which it has got right and which bits it has got wrong. Cheers
One of the uses for Gigapixel is increasing the size of an image after a significant crop. Another is for really small images you "steal" off the internet or from video stills.
One of the more interesting photo comparisons Peter. I have used Gigapixel since its first release. You have just confirmed my thoughts that I do not need to upgrade my old E-M1 to get pixel shift.
One thing I will say though is that on hot days when there is air turbulence gigapixel can give strange results with up resing of far off buildings and trees in localised heat shimmer on single shot panoramas (have not tested multi shot panos). All confirmed with pixel peeping the original shot, but not immediately apparent when viewed normally.
Good review. I used the trial Gigapixel on my E-M5 MkII images (and others) as well and found the quality of upsizing was surpringly good and the files had no problems with further processing. Was also fun taking my 3MP shots from 10 years ago and up sizing to 12 MP and they looked good as well!
That is something I need to test.
Thanks Peter. A very interesting comparison. I recently tried Topaz Sharpen AI and that worked very well.
Really enjoyed your stream with Rob the other day. Hope you do another one soon 👍
Thanks, we just might. Rob is a great guy!
I use both, I actually use Gigapixel to down size and sharping my HHHR shot, I also use denoise AI, the HHHR shot have many other advantages , such less noise, more DR, and longer exposure time.
Hi Peter, great video as usual, just a heads up, The Olympus Pen F also has the high res shot feature, Images shot at 50mp produces the same 80mp file.
I got gigapixel a few months ago and I’m amazed by how well it upscales high quality files. I’ve created 90+ mp files out of iso 100 5D mkiii files that have amazing detail and no weird artidacts as far as I can tell. Realistically, creating approximately 50mp files from
Something like R6 mkii that have as much detail and sharpness as R5 is feasible.
Great informative comparison Peter. I've been using all the Topaz Software for decades. I'm seriously considering the Olympus because of the Live Comp mode
I've been using Topaz Denoise AI as an "additional external editor" with Lightroom along with Sharpen AI. Denoise seems often to be simpler and better. Sharpen does offer rather contrasting options. In terms of image quality, I just had an old Olympus four-thirds 35-100 mm f/2.0 lens out in use.
Good video. Gigapixel AI is ideal for wildlife/bird photography where you often crop. Will make printing larger images possible.
Absolutely!
This was very interesting. I could see using this to make high res Panorama shots from my drone for sure. Thanks Peter.
Yes it would be a good software for that. I need to test that also. Have not used it with my drone shots yet.
High Peter. Nice comparison and surprising result as a high res shot does include more information than a single shot. Another thing: you forgot to mention than PEN-F does also make high res tripod (not handheld).
Thank you Peter! This was very interesting. I am tempted!
I have an Olympus Pen-F and it also has the High-Res function.
Thanks for reminding. I tend to forget about Pen-F.
@@ForsgardPeter After watching your video, I find that the High-Res function is probably best as a "scanner" or for product photos. For landscapes, Topaz Gigapixel is probably better, and then it fits with my idea of buying a Fuji GFX 50R. No need for 100S. The GFX 50R fits perfectly with the m43 cameras I already have, even the optics can both use, together with Metabone's adapters.
@@ForsgardPeter That's a pity, since that's a very nice camera which you can now get quite affordable 2nd hand. (Although I prefer the usability of OM-D and the highres shot from the Pen-F cannot deal with movement such as in water surfaces. That seems to be improved in later Olympus models.)
I was waiting for Topaz Labs discounts and then bought one, then another of their software. Eventually, I bought enough that at another discount where they offered a Suite discount, I bought the rest of the modules. The Gigapixel AI and the Denoise AI are very good for someone who does not have intense Photoshop skills or does not have the time/effort to carry out high skilled Photoshop equivalents. The enhancements do help lift some of my modern images but I was really happy with Topaz Labs software for enhancing 3 MP images from my first digital cameras, the Nikon Coolpix 775 and the later from my Olympus C-750uz and Kodak P880.
I have many of the Topaz products including AI Gigapixel. They have great products but you have to be careful with them and take a good look at the output.The latest version of Gigapixel AI can be used as a plugin for Lightroom and Photoshop. I haven't tried it yet, I assume it works. Apparently you did not do an upscale conversion but a 1:1? I have never tried that, I always use it to enhance very cropped wildlife images. Sometimes a 6:1 upscale. I get mixed results. Sometimes amazing other times obvious image issues. I haven't used the latest version yet and they do make improvements all the time. I would add that the DeNoise AI is really good for hi ISO shots with lots of noise. It does an amazing job. Anyway, thanks for your videos. Always interesting to watch.
Amazing that the Olympus is sharper and in camera. Yes, moving subjects are an issue for decision as well as corner sharpness.
Thanks, Peter. I've tried Gigapixel some time ago with good results, but ended up buying Denoise AI and Sharpen AI. My M10 mark II can now produce much cleaner and sharper images. Even my E-3 now is worthy again! I think this technology is capable of blurring the difference between high end and enthusiast /amateur equipment.
It may be interesting to compare, for example, a picture taken with an M1 with pro lens against a picture from an M10 with kit lens processed with Denoise and Sharpen.
Thanks Peter! The conclusions of your research is astonishing. What would be also interesting. DPReview use test image for all cameras they test. In the right lower corner of the image there are tubes of painting colors with letters and the letters are not discernible even in some HiRes images (if available in particular camera). It's interesting would Gigapixel applied to normal image beat performance of HiRes image judging upon the letters on the tubes? In general, test using text, paper for instance, would be more consistent than city landscape because in case of the paper text there is a strict differentiation - you either able to read the text or not. The paper has to be placed on the distance from camera such that in normal image the text is not completely recognizable or barely recognizable but not too long distance when the size of letters in pixels is too small - not less than 6 pixels height - to avoid discretization impact.
Ah, the life of a professional photographer these days... can't even afford a pair of jeans with both knees intact....... : ) This was a very interesting video.... since I only have an EM-1ii, I think I will invest in Gigapixel.... I plan to print some of my favorite landscape pictures so this will help keep them looking good. Thanks for sharing Peter!!
You should test it. The 30 day trial version works perfectly. There is enough time to test it.
Funny, I’m a professional photographer and my jeans are also ripped at the knees🥺😮😜
Look out for heat shimmer on landscapes. It may be unnoticed in the original but Gigapixel will show it up in strange ways such as a bent wall or window or a small cluster of warped trees. Having said that, its great on landscapes.
Very interesting video Peter, plus entertaining as always (love the humour upon mistakes, makes them even better:) Regards from Barcelona.
Glad you like them!
Hi Peter
Nice review and surprising result indeed. But what would be funny to check is how a high res shot could be upscaled and sharpened to let's say to 125 (handheld) or 200 (tripod) MP, because mathematically it's an evidence that a high res shot contains more information than a single shot, as it's a merge of 16 (random move) or 8 images (controled move), depending handheld or tripod. We have to remember that any camera can produce hand held high res shot (even the PEN-F 😉), if you take a burst of RAW images and merge them in post with photoshop (set minimum setting for any NR/sharpenig in RAW, open as calques, align, convert to dynamic object with median rendering, sharpen - it works well with Topaz Denoise AI which can also sharpen).
It would be interesting to test how an Olympus High res file compares to cameras with very high pixel amount.
Hi Peter, thanks for sharing this interesting tests. Is it possible to use gigapixel AI on HR files to get even better results? Not sure if it's possible. Thanks again.
Most likely it is. Didi not try how he results would out.
Peter, since Finnish people are famous for being great hockey players, you should go to a rink and take some action shots of them skating on the ice. It would be a great way to see how Olympus cameras perform in sports situations.
That could be an option. Maybe if I can find some outdoor hockey in a very cold weather.
Thanks for the comparison. Looks like the ultimate in image quality could use a combination of the two, blending in the Gigapixel result of moving objects with the stationary clarity of in-camera processing.
Sensor movement.
Thanks for the comparison. I bought DeNoise and Sharpen AI.
Good choice!
Thanks for the interesting comparison.
See you have the PD tripod, looking forward to seeing your thoughts on that. [unless I already missed that one]
Yes, it is my favorite. Have not made a video about it, but I am planning on making one.
Whau! Right on time :) I was looking on Topaz software and was wondering how good is they Gigapixel? I were planing to do my own comparison with HiRes and here you are :) with your video! And looks like is definitely worth to consider. Peter, thanks a lot for video!
Glad I could help! Yes, it is worth to consider if you need more pixels and the scene has a lot moving subjects.
Peter, after taking a High Rez shot with Olympus, did you take it a step further and process it in Topaz? Does the High Rez file from Olympus stand up for further resolution up scaling in Topaz? It would be great to know what the result was. I would like to continue with my Olympus lenses and 20mpx bodies rather than make an expensive gear change to Full Frame 40 or 45 mpx bodies.
I did not test how the high res shot upscales with Gigapixel. That would interesting to test. maybe I do a test.
I have used Gigapixel to enhanced old snapshots taken from 2009 video and it works quite well albeit a little slow (3 minutes per photo). I then finished them in Lightroom.
That would cool to pull out some stills from old VHS videos and see how Gigapixel can handle those.
My prediction: the trend of cameras is not bigger sensors, but smarter image processing technologies (e.g. AI); bigger sensors bring us less and less advantage. Smart cameras will emerge. This can be a promising future of MFT system. Some image processing software packages will be user installable, just as we install apps in smart phones. Once you install a package (e.g. something like Gigapixel AI), your camera gains a special ability. Of course, for the functions which need mechanical operations, such as driving the lens to zoom in/out, or to change the aperture, or to change the focus, we may still need firmware upgrade to support it, unless APIs are exposed to external programmers. But more and more visual effects can be achieved using pure computational approach, instead of optical approach, so less customized mechanical operations will be needed. Less firmware, more software.
Good ideas, thanks for sharing.
AI is high energy consumption thing and this is main obstacle to put it inside camera. Also AI adoption in camera contradicts trend on megapixel increase.
@@sstansm7f True. But phone cameras have already benefited a lot from AI (e.g. portrait mode, night mode, etc.) and the small sizes of lenses. If MFT does not go the same direction, it won't be able to survive between full frame and phone cameras. Regarding the energy consumption, the battery life of mirrorless cameras is shorter than DSLRs, but mirrorless still wins the market.
@@ruiduan2843 Yes, but 1) Smartphones produce low quality initial images (apprx as P&S digital compact cameras in 2004-2010) it does not require much CPU resources to enhance them 2) Smartphones enhance image on the fly as one can see them on display before taking picture 3) Smartphones does not allow serial shooting with measurements. One can add AI capabilities to cameras as post-processing via menu but the problem is that the photographer can't assess resulting image unless camera will have FHD display 6" size.
Interesting comparison Peter. I find software such as DeNoise invaluable in high-contrast situations such as sunrise and sunset where HDR isn't practical and the shadows need to be pushed in post, resulting in noise. That said, from a creative perspective I think we need to draw the line at excessive digital manipulation e.g. sky replacement or combining elements from different locations i.e. how do we define a photograph vs a digital art.
That's easy. A photograph is not manipulated. If you do add or take away objects in a photo, it's no longer a photo, but something else. Nothing wrong with that, if that's what you like. But I don't like people who pretend that a composited image is a photograph. It isn't.
Has anyone usung a hi res shot using olympus, then running through Gogapixel?
Thanks Peter for another very interesting video.
Thank you Peter! How often do you carry a tripod, and have you done a video on your favorite tripod for travel?
Not very often if I am going to photograph. For video work, I need one. My favorite tripod at the moment it is the Peak Design Travel Tripod. I have not made any videos about tripods.
'AI' upsizing is available much more cheaply within Pixelmator Pro for Mac and Pixelmator Photo for iPad. Probably just as good, too.
I just learned about Pixelmator Pro, but have never tested it.
I have both Gigapixel Ai and Pixelmator Pro and the latter comes nowhere near what the former can do when it comes to upsizing.
There are so many options and how they compare varies from photo to photo and area to area within a photo. Check this video from 17:09 to 18:30. th-cam.com/video/5LrPKoNvPPc/w-d-xo.html . You never know what error is just around the corner with these machine learning upsizing apps. That was my original point: they are inventing more information than they are keeping. Unlike the Olympus technologies that are compositing real data captured in the field from the subject. Also, if you want the ultimate in high resolution, panning is the way to go. Machine learning upsizing is a toy compare to that. If you want to play with the toys, something vastly cheaper and built into a powerful editing app like Pixelmator makes a lot more sense to me.
I setup Denoise AI, and Sharpen AI as an external editor in Lightroom, and can choose it from lightroom in the edit pull down menu.
Thanks for doing this comparison, Peter. I own Gigapixel AI, but never bothered to compare it against my E-M5 MkIII handheld hi-res. The good thing is that Topaz is constantly tweaking the AI algorithms, so its output will keep getting better. I see that Skylum has gone all-in on AI with their upcoming "Luminar AI" product. (Question: In your tests, couldn't you simply have dialed in "2.5X" and "4X" in the "Scale" setting of Gigapixel to get your 50MP and 80MP output?)
Yes I could have done that too, but I decided to use the pixel amount for more clear info.
Another advantage of Oly hi-res modes is noise reduction through averaging. Now if you really want a lot of pixels, create a tripod hi-res panorama then double it with Gigapixel AI. Btw the current version of gigapixel ai does work from Lightroom, at least on my Mac. I think Anthony Morganti did a video on that.
Ok, it does work with Lightroom. Need to test that.
My lingering question is, what is the outcome using Topaz AI in conjunction with Olympus High Resolution mode?
Thank You Chris
It depends what you are adjusting with Topaz AI.
@@ForsgardPeter MFT drew me in. I had been saying that I would skip 'full frame' and go to medium.
Results from MFT are better than I dreamed. I imagine 'full frame' is too.
I don't plan on big prints, but often crop minimalist details out other wise usable prints.
These areas can fall victim to frindging, which I have been inept at dealing with.
With my first PC in 25 years and only Windows 10 home , the search is on for software.
I do not have internet connection, so no subscriptions. Thanks Chris
The poor corner sharpness was due to the lens optical limitation. You need really good glass that is shot at optimal aperture to take full advantage of high-res shots.
Great idea- really well done video.
Thank you very much!
Interesting, always tempted to try the topaz stuff.
You should, Jimmy! Their DeNoise AI and Sharpen AI products in particular are very good. Sharpen AI can even rescue slightly out of focus shots! And it's amazing in correcting for handheld shake.
I totally agree with Carl. Topaz makes great software. DeNoise is the one I have used the most.
Thanks a lot for this rather unusual comparison...
Did you edit the HR or HHHR in Raw or simply use the JPEG?
RAW.
WooHoo..first to watch...Giga works from Lr but not Ps plugin,at least on my PC
A prize cookie for you!
@@albertocugat9689 choc' chip please.
@@garys639
Of course!
Hi Peter, thank you very much for all your hard work.
I am wondering if Topaz GP increases noise. Sorry if you mentioned it already somewhere.
Cheers!
I did not really notice that. It should be tested with a high ISO image.
@@ForsgardPeter Thank you, have a great day !
No because of the remove blur slider it produces smooth images. If you work with a real small starting image, it will look painterly, too smooth but better than nothing.
A very strange comparison, indeed! By the way Gigapixel Ai does a great job on old photos, not just right now photos! And eventually Ai will be the downfall of our style of editing.
I agree that AI will be affecting our way of editing.
Good review.But IMHO more correct is comparing for example Lightroom interpolation tool vs Topaz Ai,while both do "math" and fighting resizing artifacts ,but actually not add any new details.Olympus multi-shot mode providing true 50/80mpx and no brainer it will be better than Topaz Ai.Again IMHO.
It's amazing what image ai can do nowadays. But how often does the average photographer a high res picture? For $100,- it's expensive if you use it as a gadget.
It is no sense getting it if not needed.
It's less than $100, but it is more expensive than most of their other AI-powered apps. It's real strength is upscaling a highly cropped image back to original size, or squeezing more details from scanned photos or slides. You would rarely use it to upscale a typical 20MP photo, because there's no point (and the TIFF files created are huge!).
I don’t understand what is this gigapixel doing with the photo?
It is upscaling the photo. In this case to match the 50Mpix and 80Mpix high resolution files.
What I really need is AI for politicians in my country...
Hallo Peter!
I made some tests. If I shot with my EM-1 III in Handheld HighRes mode and do sharpening mit Topaz DeNoise AI (Yes in AICear Mode) the results are much, much better as in Topaz Gigapixel AI. Much more detail and more sharpness.
Thanks for sharing the info. I need to check to too.
Great video. Thanks.
right now: high res olympus
someday in future: AI
i actually have my fear that AI will ruin the photography once it become too powerful and smart
The big fear, and some predict (Steven Hawking, and a lot of others), that AI will ruin us and the planet. We'll end up in a battle for resources. We should nip it in the bud.
It is true and it has already done damage with Ai driven skin retouching etc.
@@ForsgardPeter Topaz Denoise AI does not rip skin textures (that would produce awful result) but emphasizes tissues of clothes.
@@sstansm7f Look at Luminar 4 or AI. You can easily do edit portrait with some sliders. Most clients wouldn't pay you for more effort when doing this with Photoshop.
That sounds like a benefit. The problem start, if all clients can do this and all the magic happens on a smartphone. By now, you mostly know which pictures are composed and what are real, especially from non photographers. But if your smartphone replace the sky, eyes, moon, sun and simulate the lightning to the background, you have to keep up against that AI composed colorful noise free pictures.
The good think is, the AI could not guide a group of persons, could not shoot a concert or wedding and is not able to help you for creative work.
@@polarized8708 Smartphones take standard pictures and capable to enhance them to the level of acceptance for average customer. It's mass product. For average people. Also smartphone made pictures are too noised and colors degrade in low light conditions far behind what cameras can produce.
I’ll have to take your word for it, because I cannot see any difference between the images on this video. I bet if you adjusted the contrast and perhaps sharpened the Gigapixel AI image ever so slightly, it would be an even closer match.. there’s no way anyone would be able to tell.
True and as I said both were straight out of camera. Maybe I will make another test and see how it goes.
Peter Forsgård I’d be interested to see it. The results are already impressive! It’s amazing what they can do with AI upscaling these days.
Peter, the 80 mpix high res raw files are always very soft and need to be downsized and sharpened to look acceptable IMO. I assume you did not downsize the 80 mpix file, however was it sharpened and if so what settings and in which program.
Great job peter thank
Thank you.
Interesting test
Great video, very useful for product photos. Thanks!
Interesting video.
I read on some website that Alien Skin Blowup is better.
Have not tested that one. I might have a look.
Good info - interesting. Maybe I can squeeze a bit more life out of my 16 mp EM1 v1 - I will give it a try🤓
Go for it!
Peter, you need a good woman to take care of your trousers.
I have a very good woman.