Topaz Gigapixel 8: What's new? Should you upgrade?

[@johnbald9639]

Just what I needed to know, thank you! Good info and good explanations.

[@ytuberization]

Would have loved to see the comparison to Photoshop ACR SuperResolution, not the comparison to the standard PS resize algorithms. Otherwise very interesting video. Thanks!

[@davidpharo]

Yeah, me too but suppose that the ‘purity’ of sticking to non-generative routes as a baseline for comparison makes sense. Also wonder just how subjective the results for any specific image and generative enhancement get -let alone when you begin to then wrangle different sliders’ levels and options into the mix!

[@onlysublime]

Photoshop ACR Super Resolution is still pretty primitive. The max it'll do is a 2X increase and there's a ton of noise that you'll have to correct yourself. And before you ask, if you want to increase the resolution again, it'll say it's already maxed out.

[@Vixpix1]

Forgot to say I enjoyed your video!

[@lennard4454]

I tried redefine on some of my own images, it's absolutely nuts!!!

[@OrigenicVFX]

Thanks, you're the first person that's mentioned the processing time, my slow laptop would take forever. I'm curious, what specs are the machine you're trying this on?

[@mk0x55]

Thanks for the review! Well, honestly, I can't imagine using AI seriously to up-res my images. The AI models just go haywire and fill in stuff that wasn't there in the image in the first place - and I think it is quite visible. Also when it comes to the traditional upscaling methods... besides perhaps trying to print a very low-res image big, I can't quite imagine where I would need an upscale like that. Even a 15 year old 5D mark 2 gives me 21 MP, which is enough to print pretty big. I've recently seen a comparison of large prints (A1 or A2... way over 16 inches) from a 12 MP camera vs. 100 MP, and interestingly, not only did several experienced photographers fail to distinguish which one was the higher or lower resoluted image; most of them actually believed that the better looking image was the one with greater resolution, which turned to be incorrect. Specific skillful sharpening for prints actually seems to make a lot of difference; and 12 MP-ish images can be blown up to pretty big prints without any real loss. Then there remains the question... what do upscalers like this actually offer and how much value is there in having that capability? Perhaps it's a thing when trying to rescue low-res images from drones, phones and such...

[@mastixmastix]

My experience with Topaz Gigapixel is really bad in that sense. Invented data and in the same image different sections have different parts of Noise Reduction and sharpening applied. I much prefer Photosho Enlarge. You don't see crips results as if it was taken with a 100MP camera but what you see is real and not invbented by AI and messed up with their bad NR

[@onlysublime]

@@mastixmastix you don't need to use AI at all with Gigapixel. And seriously? You prefer Photoshop Enlarge. it's so primitive.

[@mastixmastix]

@@onlysublime The non AI models are really bad. Just try on any image with a plain sky and you will see. It will apply NR through different parts of the image so the results is really weird with some parts with grain and some without it. And all that even if you put the NR to the minimum of 1. Horrible software IMHO. Also support is of Topaz is not good. Photoshop uprez does not do weird stuff with selective NR and sharpening. And support is top notch

[@onlysublime]

@@mastixmastix PS acr super res has crazy artifacts and noise. And 2x limit

[@mastixmastix]

@@onlysublime Nope ...maybe you are using a PS from 5 years ago. My Photoshop gets amazing result. Just the contrary of crappy results I get from Gigapixel 8

[@Ricardo-SW]

I have Topaz Photo Ai, have you had an opportunity to consider that upscaling vs the Gigapixel app?

[@onlysublime]

I have both. still hate Photo AI (preferred when Topaz Labs has them as separate software (Denoise AI, Sharpen AI, and Gigapixel AI)). So Photo AI is very similar to the basic models in Gigapixel. If you're happy with the results in Photo AI and don't need additional controls, just stay with Photo AI. But if you wish you could do more tweaking or had more models, then it's worth it to get Gigapixel.

[@mymusic5772]

Thanks Todd..Is there much difference to upscaling on gigapixel AI to using photo AI for upscaling.I,m kind of assuming you’ve tried both.

[@Rainy78]

​@@onlysublime So you can do more in Gigapixel 8 than photo AI? I dont see any deals this holiday season on photo ai. I do on gigapixel 8 though. Can I do the same in gigapixel 8 as photo ai? as far as sharpening, retouching faces, denoise etc? Thanks!

[@onlysublime]

@@Rainy78 gigapixel is for up scaling your image . Not retouching

[@joshnlevinson]

How do the standard models compare to previous versions of Gigapixel?

[@Ted-Byrne]

14:53 Nice review, thank you. GP8 is on sale now…. I was looking for the best 2x-3x upscaling option for simply making larger prints of relatively recent digital images. Not caring at all about the generative model and AI stuff, it looks like the standard resize model will do it for me. 👍🏻

[@OnlyThomasHayes]

This video was right on time. Also pleasantly surprised to see that I am already subscribed to you. Do you think paying the renewal fee for automatic upgrades is worth it?

[@abbassepahvand8819]

Hello! Todd, I just wanted to say thank you so much for your videos, and especially for their high quality. I have a Canon EOS Mark II and a Sigma 24-70 f/2.8 lens that I use for my KZbin videos. I also have advanced lighting equipment and recently bought your custom Log LUTs. However, my video quality doesn’t quite match yours. Could you please share your camera settings for recording videos like this one? For example, what framerate and ISO do you use, do you use a color checker, are you recording in C-Log3, and what autofocus settings do you use for your studio videos (not for outdoor recording)?

[@MarkOvergaard]

Thanks for the review, Todd. My instincts are also to stick with the basic models. Do you have any impressions regarding the benefits of the high res model vs the standard model, especially for images captured by cameras in the R5 class (45Mp)? At least initially, I'm not actually upscaling from 1x. My main focus is using face recovery to (hopefully!) reduce motion blur effects on the faces of dancers in low light contexts where even 12800 ISO w/ f/2.8 won't get me above 150-250/sec shutter speeds. I've just started experimenting and it seemed like there was a modest win with the high res model. Any suggestions from you or other folks in this group?

[@josdenis3684]

Great review! Thank you.

[@MaartenOosterbaan]

Thnx for the comparisons, what I kinda missed is...are these non AI models of version 8 much better than in version 7 or 6

[@pacomolinari]

Hi Todd, just want to say that I really enjoy your videos! Just a question about upscaling: what about Topaz Photo AI? How does it perform against Gigapixel? Is it worth buying if I already have Photo AI? Thank you!

[@elhp3818]

Thanks for the review

[@onlysublime]

Long-time user of Gigapixel and I love it. A little wary of where Topaz Labs's direction is going but still a customer. I took a 1792x1024 anime image and 4x it using the Recover generative AI model. It took almost exactly 3 minutes to finish 1 image using my PC with Nvidia 4090 GPU. It popped up a box saying that the Cloud Render would take 3 minutes and 49 seconds and would cost 12 credits to render. So it would take longer to do it in the cloud and cost credits. I then took the same 1792x1024 image and 4x it using the Art & CG basic model and it took 3 seconds to render and save. So comparing the images, I overall kind of preferred the Recover model. But it rendered colored regions as if it was a slightly textured painting style as opposed to the more accurate solid color of the Art & CG basic model version. Was it worth going from 3 seconds to 3 minutes? Not really but at least now I have the option for other types of images. I then took the same image and used the Redefine BETA generative model and it took 3:05 using my GPU. This time, it didn't offer me a cloud time estimate. Photoshop still hasn't caught up yet to Gigapixel.

[@mastixmastix]

Results from Gigapixel 8 are really bad. Patches of different NR in the image give beyond horrible results. Amazing that people let their images being destroyed by this subpar software

[@onlysublime]

@mastixmastix well there are two ways in the workflow. If it's a noisy image you should denoise first. I use denoise ai for that. Or you can use the masking tool. It's still better than what Adobe camera raw offers

[@mastixmastix]

@@onlysublime Well we disagree on that. I have tested quite a few images and Topaz always looses out to Adobe and the results are always fake looking and unnatural IMHP of view of course.

[@onlysublime]

@@mastixmastix yeah totally disagree. Adobe is too limited and too noisy

[@stevencasteel]

LOVE that they have finally added creative upscale / generative AI to their powerful local toolset. Magnific is so stinkin' expensive. And even though Leonardo and Krea and Midjourney are reasonably priced, they are still a monthly fee.

[@SteveVickAK]

You sold me on it! Now I don't have to buy a 50megapixel camera just for that random time I want to print larger than 16x20.

[@ashleybirch93]

whats your take on gigapixel vs their photo ai for upscaling?

[@ThePNWRiderWA]

I have it installed and played with it on my M1 mini and I have 16 gigs of memory in it. It took about a day and a half to do the generative version. I do understand why topaz came up with the point system or whatever they’re calling it doing it on your computer takes a long time a very long time, but I can just trade in my Mac one mini for one of the M4’s loaded up with memory and see how long it takes I don’t usually do videos. I used to have a channel at the infancy of KZbin and I was associated with the product and I had to surrender all the videos I had done to the company when I left, but I think I am going to start making a few more One of them that might be interesting would be a comparison of doing the same image on gigapixel 8 with each system since I did order the new mini. Does it really make a difference with the ARM chip and more memory ? Maybe the bottleneck is the video card and if i used a Pc with a powerhouse video card GPU it would make majors of a difference ? Where is the heavy lifting being done ? My interest is I am nearly 69. I have bad knees and much of the time I use a MFT OM-1 with the 15-400 lens with wildlife. My Sony A7 IV with the zoom that goes to 600 is just too big for this old man Having a way to get better sunrise and sunset wildlife images on MFT is welcome and it has done a good job for me

[@onlysublime]

if you had a faster GPU (I have an Nvidia 4090), you can render pretty quickly. Just a few minutes. M1 is great but it still didn't have a great GPU.

[@ZachGroney]

Question: Has the basic standard model changed from, lets say version 6 to version 8?

[@alexlford]

Thanks Todd - would be good to see a video comparing high megapixel cameras, Lightroom enhance and gigapixel 👍

[@lazlo342]

Perhaps I missed where you said it...what model GPU are using?

[@spydergs07]

I run a 4060 and it runs Gigapixel 8 amazingly.

[@alankefauver6187]

I don't seem to have the option to render in the cloud. i am on a PC. That button is not next to the "Export to Lightroom Classic" Reinstalled and still no. V 8.0.0

[@ytuberization]

Credits for unpredictive outcome? Mmmhhh…🧐

[@spydergs07]

Just get a decent GPU and you can run it on your local machine no prob

[@ytuberization]

@ Is the result identical?

[@FotomakerAcademy]

​@@spydergs07Not so. Even NVIDIA 4090 isn't fast. No one below that can run locally.

[@spydergs07]

@@ytuberization Running a good GPU will run much much faster.

[@spydergs07]

@FotomakerAcademy I run an I7 14th gen 20core cpu, 64gb Ram, and 4060 GPU with 8GB ram and I'm able to run Gigapixel 8 great on my machine. Even running the beta recover and stuff snd exporting only takes 2-4 mins tops. I actively use it daily on my machine and have Zero issues. A 4090 would smoke my 4060 also.

[@FotomakerAcademy]

I got debited 12 credits for a Cloud render and it made zero alteration to my image. So a large waste of credits for no result. Did you run a Text Preserve with Hotel Keeler text?

[@NoahtheGameplayer]

would a shadow PC work honestly? I mean yes i pay 50 dollars a month to Use the Cloud PC, However, at least its faster and cheaper then buying the credits, so.... what would you guys think?

[@Jeannemarre]

I am still on v7 and the times to render when using recovery are still Soooo slow, and I run an M3 MAX MacBook Pro… it seems like it’s still not optimized for MacOS

[@moserphotoworks]

I decided to upgrade from version 7 after catching a KZbin review that showcased the new generative models-really impressive results on display. But when I tried it on my own setup, it was a huge letdown. Even using the smallest image, I couldn’t get past the preview to do anything remotely creative. It’s just not usable. I haven't tried the cloud-based option yet, but honestly, I’m not looking for another pay-as-you-go setup just to process images. The Basic models here are fantastic, but the generative models feel half-baked-Topaz still has some work to do.

[@spydergs07]

What specs do you have in your PC?

[@DaveBowman]

If I’m understanding this correctly, the Generative model doesn’t just use Ai to improve the detail across the image I.e. add in detail where required, but it actually creates a completely new image using the base image as reference. If that’s right, then don’t you just end up with an Ai image which is technically no longer a photograph? Ergo any Ai checking s/w is going to flag it, and it may also move copyright into a somewhat grey area.

[@PowerRedBullTypology]

Well, but since you have the original and maybe you uploaded it before somewhere, the AI version is basicqally a copy of yours and thus the AI does not have the copyright itslef, but i would guess that any judge would say if the AI is htat close to your original, the AI one is stolen from the original. However, since you cna not steal from yourself, the other party who might use it would break copyright cause you have the original that is very, very close?

[@RockTo11]

Photoshop has a few other better quality upscaling options. One is inside Camera Raw, called "Enhance". Another is in the Neural Filters called Super Zoom.

[@spydergs07]

I like Photoshops upscale options, but Gigapixels upscale is better.

[@onlysublime]

Enhance is very limited. it's limited to 2X increase and produces a lot of noise and artifacts. Super Zoom also has a lot of limitations and has the same noise/artifacting problems. it also requires that you have to manipulate what you want to zoom and because it's slow that you can't really experiment.

[@johnritchie6775]

I made a comment last week about the slow processing even with a fast PC & how long I was waiting so I just want to update everyone how the processing went…err I’m still waiting for it to process 😉

[@spydergs07]

What specs do you have on your PC?

[@Hanneth]

The processing id dependent on your GPU. The better the GPU you have, the less time it takes. Video AI I know you can force to use CPU instead, which I had to do for some videos on my old GPU. I've upgraded and no longer have a problem. I'm not sure if in Gigapixel and Photo AI you can force CPU rendering. CPU rendering is a lot slower and should only be used if GPU rendering will not work.

[@mymusic5772]

I have just purchased an M4 MacBook Pro with 24mb ram.Would anyone in the comments know if that would run this software well or would it be considered too slow.

[@Hanneth]

@@mymusic5772 I was running this on an AMD Radeon RX480 and it worked well enough for me. I've upgraded since, and the new LLM models take a while to process on my Radeon RX6800, but the other models are quite snappy. For the redefine, and refocus you are probably looking at minutes for final output. I think even rending a test section can take a minute, or 2. For the Apple Silicon chips I found it uses the neuro processors. The results customers are getting suggests the M4 is slighly better than a Nvidia GeForce RTX 4060. My card puts me at the equivalent of not quite a RTX 4080. That would suggest that a M4 would probably take about 5 minutes to render a preview of refocus, or redefine. Maybe 15 to 20 minutes to process the whole image. The more traditional Low Res, Standard, High Fidelity, Text & Shapes and ART & CG would probably be quite snappy. Depending on what you are doing, different models will work better for you. Sometimes I end up blending an image together using different upscaling models. Face AI should also work quite well with a M$ as well. My understanding is that uses a form of LLM. What I would do is download the trial and try it. If you can live with however fast your computer can render things, this product is really worth it. If you are working more on actual photographs, I would recommend looking at Photo AI instead, which can denoise, sharpen and remove objects as well. I personally use both.

[@jessejayphotography]

Not a fan of the generative AI features. Like, WTF is that. Your example with the old photograph of Hotel Keller is a prime example. Dangerous tool in my opinion, especially in the hands of someone not well schooled in the AI stuff.

[@Hanneth]

It is important to understand that Gigapixel AI isn't just for realistic images, never has been. It is for upscaling any type of image. This now includes the ability to become creative with upscaling the images. If you are just looking for realistic image upscaling, that is what Photo AI is for. Topaz Labs has stated that they will not bring all the features of Gigapixel AI to Photo AI as they serve different purposes. Gigapixel is a general upscaling program, that can upscale faster than Photo AI as it can optimize things better without having to worry about the other AI tools. Photo AI is designed for Photos. I use both. I have found some of the new features in Gigapixel helpful at times in upscaling realistic photos, but mostly use it for cartoons and other more art based projects. The redefine I find especially interesting as Topaz Labs uses making AI art look more realistic as an example of what can be done with out. I think they've also shown using it to make AI art look more like art.

[@pixeljauntvr7774]

What on earth are you talking about? "Dangerous tool" You have the same mentality with spell checks or calculators? If humanity thought the way you do today we'd have never made it past the wheel.

[@jessejayphotography]

@ your comparisons are ludicrously simple. Did you watch the video? You want AI enhanced documentation photos changing fundamental elements in a photography? You want AI changing aspects of lettering on your clothing in a family picture?

[@KaosK9]

yeah... My image was 20 credits for a 15 min cloud render. Took 30 mins local.... What a joke... Finally the apps are starting to do what they claim halfway reliably, and now they want more $ from your pocket.

[@TheShepherdsChapelChannel]

Topaz needs to invest in improving the software. i feel they been static for a long time no real leaps for better images

[@thewebstylist]

Already an expensive program then only allowing 25 cloud renders before charging more, wow 15 mins to upscale local is ridiculous

[@OllieOtterOxen]

Does it use gpu acceleration for local uscaling?

[@chrisstocker]

I’m trying to upscale a picture and it’s taken 12 hours so far it’s absolutely ridiculously slow, it’s unusable

[@Vixpix1]

I do not like Gigapixel 8. I am about to uninstall and go back to the last updated version of Giga 7. It does not come close to cleaning and upscaling my images as the previous version did!!!

[@mortenthorpe]

Paid for every version of it, hardly ever used it with success… pulling the plug on it

[@graham_T]

Recover and redefine are not yet fit for purpose and I am in discussion with Topaz . Furthermore the price model they show per mb is not being applies properly and every time I am overcharged.

[@marcusavanti902]

I have Topaz Photo Ai, which is much more expensive than Gigapixel and was meant to replace it but the results for pictures with faces, humans are just terrible, unusable. I litterally don't use it. The marketing, pr videos make it look amazing but I have never seen anything like the results they show.

[@fabiankeppler315]

down the rabbit hole... now you can still use your own (not so powerful computer, as if) ... hey if Adobe can do it.. why not topaz

[@G.Menounos]

Not a fan Of having to buy credits to use an item of software. A monthly or yearly plan might change my mind, but not a per credit software licence. It limits the images i wanna to retouch and my mood.

[@Hanneth]

Topaz Labs remains committed to allow you to run all AI models on your local machine. They also plan to work on trying to optimize the new models to run faster over time. If your subscription to their upgrades runs out, you can always use the last version available to you for as long as you want. Local processing of everything is one of their main tenants. I personally used Gigapixel for a year without upgrades. There was demand for cloud processing from businesses, so they added that ability. You can buy credits, or there are montly and yearly plans. (I do not know the details on the plans as I am not on my home computer) In introducing the new cloud feature, they have wanted to make it very clear that there will be no features developed that require cloud processing. The ability to local process everything is one of their competitive advantages they do not want to give up.

[@Dilandau88]

How on earth does Photoshop not have this yet? Its super resolution is completely weak and inferior.

[@lhdz12]

The rendering of text is still crap...waste of my money.....does a good job on images without text....."NOT READY FOR PRIMETIME"

[@marcgilles4520]

Skip the photoshoot.

[@TheShepherdsChapelChannel]

cloud render, slippery slope to being a Adobe like problem.