Ok, i NEED to point out some stuff: 1. The unfocused video of the M50 is still visibly out of focus, maybe a little more sharp but still out of focus. In all my experience with topaz i found out that no matter how much you try to make an out of focus shot look good it's simply impossible. The model does not fix focus, it fixes details, so if you take an out of focus image at 720p and make it 4K it will always look like an out of focus 4k shot. The Topaz Video AI enhancement tab does not fix shooting mistakes or lens mistakes, it fixes file size and compression thus making the image look "sharper" 2. The non gimbal stabilized footage looks great BUT it won't stabilize a shot as good as premiere or davinci if it's too jittery and even with little movement it can show abberations 3. You need to respect the 180 degree rule or at least shoot at fast shutter speeds if you know you footage will be slowed down, or if there's a small chance it will be slowed down, because it will simply look blurry if you shot you footage at 1/50 and slowed it to 120 or 240 fps. No software can fix the insane ammount of motion blur a 1/50 shot at 120fps has. Great video tho !
A supercomputer should just upscale old movies and shows at max quality 24/7, it’s so easy, and costs nothing but power, and a person who can just make a few clicks every couple of days or weeks, depending on how much content you can queue up for rendering.
@@chclonyBoth are used I have set to GPU, and cpu is most of time near 100% when my graphic card often not fully used. Depends on used model. Overall I can say cpu is very important and heavy used. The graphics card is of course still important, if you have a bad one rendering in Topaz is very slow.
I use Topaz often, but the stabilization definitely isn't great. Been getting better results with Gyroflow and other stabilization software. Definitely not gonna replace the gimbal, you can even see the warpiness on the background from time to time.. But it's still an essential tool to fix even small compression issues that you can have when exporting from an NLE.
I think concerns over Topaz Labs’ Video AI being responsible for film-maker’s breaching copyright laws are nonsense. To be sure Topaz Lab’s Photo AI can seemingly make use of its own pre-sampled faces to enhance an edited human image. However this, along with whole image generation is not the case with Topaz Labs’ Video AI.
The thing is the amount of time you save by simply getting the shot right with the gimbal on the spot can't be matched by post if you are looking into saving time. Or for any shot for this matter, Topaz should be used to save a shot you took that you though looked good, but didn't and you find that it needs correction. Tt should not be used as an "I can photoshop that" type of tool, unless you don't value your time IMO.
Well the samples don't say much. Can achieve the same with AviSynth or VaporSynth script or even in Davinci Resolve. Should have chosen samples that are more complicated.
about june or july i downloaded the topaz demo. I was so impressed with it I wanted it immediately. I waited til their black friday sale instead. I have some old videos from 10-15 years ago that were recorded 480p 15 fps, and after topaz, if you look now it's1080p 60fps, as if I used a regular modern cell phone to shoot them. It's literally like magic, i never get tired of it. I'm redoing my entire backlog of home videos and reuploading. The difference in quality is just staggering.
The advantage of a gimbal over post stabilization is your ability to use focal length like 50mm without much background information and still have a smooth shot. Unless using a camera with gyroscope stabilization capabilities you need to see the background to smooth it
I use Topaz to upscale old videos that are originally 480-720p to 1440p-4k. I justvwant to enjoy older videos with higher quality. I never share them as my own work.
I'm pretty sure a cheap rig would be more stable than a handheld. Handheld cameras are a lot more wobbly than you might think. Even a basic rig can stabilize most of the major issues.
The only downside of Topaz Video is the processing time. I have a fairly high end PC and graphics card and it still will take around 8 hours plus to upscale a 1 hour hour video.
How would the software fare with upscaling 360p or 480p to 1080? for old videos I shot in the 80s and 90s? Those I really want to upgrade and conserve for posterity in good condition. Can it actually do a good job?
Do filmers intentionally shoot in a wider than needed angle, to anticipate post stabilization? Framing an image too perfectly in real time gets it cropped too much, I imagine? Can cameras project a smaller frame on the view finder, customized by filmer?
I tried to upscale 1min video 2.7k to 4k. On i5 11th gtx3050ti 48Go Ram. It took 12 hours to render. Can you share best setting or best entry level computer that can handel topaz
I think the first example could have come out better but it looks to me like the shot is focused on the wall behind. I'll have to convice my work to buy a license
Hey man, hopefully you see this. Can we talk about your work flow? Like would you record this in SLOG and first go color grade it and then bring it over to this and then back to a sequence in premiere or davinci whatever you use? I'm thinking along the lines of I shot in 24fps but want to make it slomo in a sequence I am building. Not sure the best way to go about doing it.
10:47 Overall Topaz is great software, but this "Mosaic" images are sellers examples! The contrast is greatest here and there is no noise, the software can calculate the gradations between this clean pixels and shows you a dream result, but in reallity this situation is rare
No problem, did you learn to do that yourself or..? Or did you have people do it for you, I am myself still learning editing videos and I cant imagine the time it took to make something like this. @@unstableworldai
This is great! So you had lot of practice. Is film shcool expensive? I would love to find a premiere pro project containing all of this editing so i can imitate it lol. Do you think I can pay you to obtain the video premire pro "project"? haha @@unstableworldai
If verticals and horizon line are important (real estate - architecture) - maybe gimbal the better choice. That said, the software is impressive - the ability to use a lower frame rate in lower lighting scenarios and still get that quality of slow-mo is maybe worth the price
Honestly, I’m all for it. Tools are exactly that, tools. Use them or not, but if in can enhance and improve your work then fantastic. I shoot regularly and this won’t eliminate my work, or the skills needed to produce. It will add and compliment to it. These same anti tech arguments happened during the transition between film and digital. Or when photoshop first emerged, it was also considered “cheating” yet it enhanced our work. This is a argument that happened during every new transition in technology
Footage that went through AI software for these enhancements cannot be classified as AI generated since the footage was not AI generated. You wouldn't classify photos touched up by Photoshop as AI generated.
I had no luck with enhancing old videos. It always looked bad no matter what i tried. In my opinion topaz can make ok footage better but it does nothing for 480 p videos from the past. The slomo feature is very good tough.
Its really good software for repairing footage. I know a lot of fans of really old shows are using Topaz Video Enhance A.i. to upscale their favorite old shows. A lot of these old shows may never get a HD scan and its even more of a issue with Sci-Fi shows as they would record the film back to video tape and then do their special effects on computers. So those shows need to be rescanned and the special effects need to be redone.
Original poster here on my other channel -- Nope, I actually used some footage I took of her for some of the testing 😅 - th-cam.com/video/kYaObVxe95o/w-d-xo.htmlsi=mjTT93jbEEMhk7zG
The fact that Topaz don't do dust or scratch removal is disappointing. Can't find any sotware that can do a great automated job with that. Non that work for me at least.
AI has is drawback for sure. But having the possibility improve the resctiction of my gear is really neat. I don't need a 10k camera to have a good low light movie, my APSC will do the job even without a gimbal. This is so convenient ! And with or without AI, a boring video will stays boring.
I've used Topaz before, and other AI based video software. It's quite impressive how far AI has come. I'm actually specializing in AI now, but I've got a ways to go before I will fully understand the mechanisms that enable this sort of stuff.
Keep in mind that Autofocus on any camera is trained the same way as AI and usually can only focus on a face or eye. So technically, if you use a Canon autofocus (or whoever) instead of pulling focus manually, you are using ancient AI. But in cine work autofocus is too unpredictable (especially when its trained subject leaves the frame and enter again) and when it does work its too digital and unnatural. So far lidar based focusing seems to be moving in the right direction so ethically I think it comes down to what is real and what is not. Command line or computer/fantasy invention of a photo or video is NOT real and may be unethical on a case by case basis and will eventually implode if it is left to its own course. This is why. I call it the Red Duck postulate. Suppose you and 2 million friends did a command line invention of a Red Duck and all of you got a fantastic image of a red duck in return. Now suppose that the global compendium of duck images (say brown ducks) totaled 1 million duck images and in a matter of a few weeks your friends uploaded your two million "photos" of red ducks onto the net. AI training itself on ducks would begin to think red ducks are natural when they are only AI inventions. The model then begins to collapse on itself because its been fooled by 3 million ducks of which only 1 million are real. So the AI training model collapses into unreality or better termed an image idiocracy. Furthermore small children will be told by AI that red ducks are real when they are not as the line between reality and fantasy is blurred beyond hope of recovery causing a backlash by the "organics" i.e people that have seen a lot of ducks in nature but never seeing a red duck do not believe they exist despite what AI now tells people.
Eventually, the quality of AI will not be affecting by learning on self collected media vs public/paid content. With the software getting better, it will need less content to learn from. Eventually, a big AI company could suffice with a tiny studio collecting exactly the kinds of source media to train its AI off. Disagree? Please tell me so I can learn.
from what I understand, its a good software, but good equipment just cannot be undermined, the enhancement seems to be laggy which probably can be fixed but that's extra work, and when it comes to stabilization, you can't compare anything to a gimbal, though I like topaz over warp stabilizer, rather rent a gimbal or a better cam for a job and use this software to enhance it further if needed
And.... it looks a *little* bit better. Let's not forget that the best technology in the world still cannot compensate for poor film-making. Same with audio production... something I've dabbled with my whole life... despite improved access to higher end tech... 90% of what I do is still crap.
i can never see the differennce no matter what settings i pick .even this guy who knows what he is doing i can barely see the differences in this video. im not sure its worth leaving your pc runnng at 100% for 5 hours just to very very marginally improve a 10 minute clip.
Difficult to take this seriously when the thumbnail is fake, having just blurred the original image for the "before" image, and is unsurprisingly not in the video. "WTF" indeed.
That "iris model 2 passes" looks super fakish and is totally unusable in my opinion. On the slomo side there are no wishes open for me on davinci resolve, which is doing a awesome job.
I see what you mean, I think you can get better results than I did with some more messing with the settings, plus the model keeps improving. New Iris model just came out as well.
Good review and good point about video "technique" being really computer "power abilities" but should we all consider how many current very succesful singers and music stars are just a result of autotune trickery? No one cares about that as they sound great, so why should we care about video lookiong perfect ? Isn't that what we want when not having big producer's resources ? Greetings from Paraguay.
People who throw ALL A.I into the same category of “stealing art” are ridiculous. Like please show me where the “stolen art” exists within he final result? It’s only referencing other videos to understand what hair detail skin detail and eye detail looks like. It’s like me asking someone to draw a man with blond hair and blue eyes then being upset that they are using memories of people with blond hair and blue eyes to know what blue eyes look like and what blond hair looks like even though the end result looks nothing like any of the references they have seen. Some A.I models are terrible and really just an image compositing algorithm where you say “Santa riding on a plane” so it finds an image of Santa and finds an image of a plane and does a quick and rough composite of both images. That’s NOT the A.I most people use most algorithms if you put the same prompt will find 3,000 images of Santa all by different artists then 3,000+ images of different planes and will recognize the similarities between all 3,000 images to understand what Santa looks like then it will rebuild a version of Santa based on that. Can’t say “BUT HE’S WEARING RED!” Yeah 99.9% of Santa’s are. Tell an artist to draw Santa and they’ll be doing the same thing.
the more similar an AI works like a human, the more obvious should be that complaining humans complain about themself. this also means: the more the AI generates photos, videos and stuff the more the AI learns from its own generated stuff ...
when filming documentary verite there are many times that filming something over again is impossible. This has amazing implications for doc filmmakers that might have otherwise thrown out an entire part of their story due to a technical error which happens when in the field following subjects for hours on end.
funny how Topaz Video AI didn't do shit for my videos I wanted to enhance, it actually made them worse. Other AI alternatives were even worse and apps crashed for no reason. Topaz destroyed the sound, So I had to recombine video and sound after. Not a good experience at all.
I think you had a bugy version, Topaz pass trough audio (muxing) without conversion, works very well. Video depends on source material, choosen AI model and parameters. For (cinema) movies, keeping max naturelness, use "GAIA" or "Theia - fine tune" model. There are others cases where Iris and Proteus are not a good choice. I think Topaz should focus here more what you can do when "reconstruction" of something fails, respectively a model that can handel better such sources.
Ok, i NEED to point out some stuff:
1. The unfocused video of the M50 is still visibly out of focus, maybe a little more sharp but still out of focus. In all my experience with topaz i found out that no matter how much you try to make an out of focus shot look good it's simply impossible. The model does not fix focus, it fixes details, so if you take an out of focus image at 720p and make it 4K it will always look like an out of focus 4k shot. The Topaz Video AI enhancement tab does not fix shooting mistakes or lens mistakes, it fixes file size and compression thus making the image look "sharper"
2. The non gimbal stabilized footage looks great BUT it won't stabilize a shot as good as premiere or davinci if it's too jittery and even with little movement it can show abberations
3. You need to respect the 180 degree rule or at least shoot at fast shutter speeds if you know you footage will be slowed down, or if there's a small chance it will be slowed down, because it will simply look blurry if you shot you footage at 1/50 and slowed it to 120 or 240 fps. No software can fix the insane ammount of motion blur a 1/50 shot at 120fps has.
Great video tho !
Literally upscaled and quality enhanced an entire 2 hour movie to 4k and it looked insane. Took 9 hours to render tho 💀
A supercomputer should just upscale old movies and shows at max quality 24/7, it’s so easy, and costs nothing but power, and a person who can just make a few clicks every couple of days or weeks, depending on how much content you can queue up for rendering.
most models in Topaz are more CPU hungry than GPU, for example Ryzen9 79xx is your friend 😉
Wow...more impressive is that it didn't crash, did it?
@6880 Nope, app works much faster using GPU instead of CPU. You can test it out changing settings in processing preferences.
@@chclonyBoth are used I have set to GPU, and cpu is most of time near 100% when my graphic card often not fully used. Depends on used model. Overall I can say cpu is very important and heavy used. The graphics card is of course still important, if you have a bad one rendering in Topaz is very slow.
This was the point of technology, to progress humanity making everything easier to do.
@3:20 the bottom lip looks very weird it added very shiny bright ridges and they sort of move around instead of staying put.
I use Topaz often, but the stabilization definitely isn't great. Been getting better results with Gyroflow and other stabilization software. Definitely not gonna replace the gimbal, you can even see the warpiness on the background from time to time..
But it's still an essential tool to fix even small compression issues that you can have when exporting from an NLE.
I think concerns over Topaz Labs’ Video AI being responsible for film-maker’s breaching copyright laws are nonsense. To be sure Topaz Lab’s Photo AI can seemingly make use of its own pre-sampled faces to enhance an edited human image. However this, along with whole image generation is not the case with Topaz Labs’ Video AI.
The thing is the amount of time you save by simply getting the shot right with the gimbal on the spot can't be matched by post if you are looking into saving time. Or for any shot for this matter, Topaz should be used to save a shot you took that you though looked good, but didn't and you find that it needs correction. Tt should not be used as an "I can photoshop that" type of tool, unless you don't value your time IMO.
Well the samples don't say much. Can achieve the same with AviSynth or VaporSynth script or even in Davinci Resolve.
Should have chosen samples that are more complicated.
about june or july i downloaded the topaz demo. I was so impressed with it I wanted it immediately. I waited til their black friday sale instead. I have some old videos from 10-15 years ago that were recorded 480p 15 fps, and after topaz, if you look now it's1080p 60fps, as if I used a regular modern cell phone to shoot them. It's literally like magic, i never get tired of it. I'm redoing my entire backlog of home videos and reuploading. The difference in quality is just staggering.
it's so hard to find documentation on which presets to use and parameters
It's because every video is different. It simply takes experience and a lot of trial and error.
any free options/ alternatives?
The advantage of a gimbal over post stabilization is your ability to use focal length like 50mm without much background information and still have a smooth shot. Unless using a camera with gyroscope stabilization capabilities you need to see the background to smooth it
I use Topaz to upscale old videos that are originally 480-720p to 1440p-4k. I justvwant to enjoy older videos with higher quality.
I never share them as my own work.
I'm pretty sure a cheap rig would be more stable than a handheld. Handheld cameras are a lot more wobbly than you might think. Even a basic rig can stabilize most of the major issues.
I have an out of focus shot from an ARRI Amira UHD ProRes 422HQ. If I run it through Topaz, will I still retain the metadata?
How strong of a computer do you need to use this software? It seems like it needs a VERY strong computer for all the calculations on each frame.
Could you do a video on topaz video ai stabilization vs mercalli 6 stabilization?
That slow mo feature was a thing of Sci Fi just two years ago. People would say it's a blasphemy to shoot in 30fps and slow that to 120fps.
The only downside of Topaz Video is the processing time. I have a fairly high end PC and graphics card and it still will take around 8 hours plus to upscale a 1 hour hour video.
How would the software fare with upscaling 360p or 480p to 1080? for old videos I shot in the 80s and 90s? Those I really want to upgrade and conserve for posterity in good condition. Can it actually do a good job?
Do filmers intentionally shoot in a wider than needed angle, to anticipate post stabilization? Framing an image too perfectly in real time gets it cropped too much, I imagine? Can cameras project a smaller frame on the view finder, customized by filmer?
I tried to upscale 1min video 2.7k to 4k. On i5 11th gtx3050ti 48Go Ram. It took 12 hours to render. Can you share best setting or best entry level computer that can handel topaz
I have this and if you have HD video, it will look a little better if waxy, forget 360 the stuff that really needs rescuing.
I think the first example could have come out better but it looks to me like the shot is focused on the wall behind. I'll have to convice my work to buy a license
Yep, that's why it's impressive. The focus was on the wall, yet I was able to recover the footage somewhat.
@@unstableworldai Oh right., Maybe I skipped though too quick. I thought you were just talking about an upscale of resolution
Hey man, hopefully you see this. Can we talk about your work flow? Like would you record this in SLOG and first go color grade it and then bring it over to this and then back to a sequence in premiere or davinci whatever you use? I'm thinking along the lines of I shot in 24fps but want to make it slomo in a sequence I am building. Not sure the best way to go about doing it.
I would color grade it first and then use topaz labs. I would also only cut the part that I need slow moed for Topaz Labs to save time.
10:47 Overall Topaz is great software, but this "Mosaic" images are sellers examples! The contrast is greatest here and there is no noise, the software can calculate the gradations between this clean pixels and shows you a dream result, but in reallity this situation is rare
OK just finished the video, the quality of your video editins is INSANE man.
Appreciate it!
No problem, did you learn to do that yourself or..? Or did you have people do it for you, I am myself still learning editing videos and I cant imagine the time it took to make something like this. @@unstableworldai
@@DiamondsInterNet i went to film school actually haha, i edited all this on my own
This is great! So you had lot of practice. Is film shcool expensive? I would love to find a premiere pro project containing all of this editing so i can imitate it lol. Do you think I can pay you to obtain the video premire pro "project"? haha
@@unstableworldai
If verticals and horizon line are important (real estate - architecture) - maybe gimbal the better choice. That said, the software is impressive - the ability to use a lower frame rate in lower lighting scenarios and still get that quality of slow-mo is maybe worth the price
Great video Kevin, this software is insane!
Thanks man!
Honestly, I’m all for it. Tools are exactly that, tools. Use them or not, but if in can enhance and improve your work then fantastic. I shoot regularly and this won’t eliminate my work, or the skills needed to produce. It will add and compliment to it. These same anti tech arguments happened during the transition between film and digital. Or when photoshop first emerged, it was also considered “cheating” yet it enhanced our work. This is a argument that happened during every new transition in technology
Read Heidegger's 'Question Concerning Technology'. Tools are not just tools.
The video editing of this channel is incredible, you only have 134 subs? Count me in anyway!
Footage that went through AI software for these enhancements cannot be classified as AI generated since the footage was not AI generated. You wouldn't classify photos touched up by Photoshop as AI generated.
If you buy topaz enhance do you get free upgrades for life?
I had no luck with enhancing old videos. It always looked bad no matter what i tried. In my opinion topaz can make ok footage better but it does nothing for 480 p videos from the past. The slomo feature is very good tough.
topaz just hit the top not only in video editing but also in image editing
My good sir where can I install this program 💗
Topazlabs.com
thanks for the breakdown!
Its really good software for repairing footage. I know a lot of fans of really old shows are using Topaz Video Enhance A.i. to upscale their favorite old shows. A lot of these old shows may never get a HD scan and its even more of a issue with Sci-Fi shows as they would record the film back to video tape and then do their special effects on computers. So those shows need to be rescanned and the special effects need to be redone.
At 5:18 the gimbal shot is cropped alot more than the A.I version..and in my opinion the A.I version wins hands down...
Great Video!
loved the beat drop at 4:43 😍
Making my work look better is not undoing my work. It's enhancing it. If I use After Effects, it's still my work.
Casual out of focus is part of the real good cinema.
Great video bro
Is the presenter Ai? I mean with all the weird "Dizang, that's daft dope" strange American lingo, not working...
Lol, I just notived that only your head is moving in the first few shots.
Hey I recognize Elle Lee from the thumbnail. I think you tried to sneak it in there, didn't you? 😂
Original poster here on my other channel -- Nope, I actually used some footage I took of her for some of the testing 😅 - th-cam.com/video/kYaObVxe95o/w-d-xo.htmlsi=mjTT93jbEEMhk7zG
0:54 the timing of that neck slide lol
The fact that Topaz don't do dust or scratch removal is disappointing. Can't find any sotware that can do a great automated job with that. Non that work for me at least.
AI has is drawback for sure. But having the possibility improve the resctiction of my gear is really neat. I don't need a 10k camera to have a good low light movie, my APSC will do the job even without a gimbal. This is so convenient !
And with or without AI, a boring video will stays boring.
1:03 “an experienced gimbal operator” 🤣
I've used Topaz before, and other AI based video software. It's quite impressive how far AI has come. I'm actually specializing in AI now, but I've got a ways to go before I will fully understand the mechanisms that enable this sort of stuff.
Keep in mind that Autofocus on any camera is trained the same way as AI and usually can only focus on a face or eye. So technically, if you use a Canon autofocus (or whoever) instead of pulling focus manually, you are using ancient AI. But in cine work autofocus is too unpredictable (especially when its trained subject leaves the frame and enter again) and when it does work its too digital and unnatural. So far lidar based focusing seems to be moving in the right direction so ethically I think it comes down to what is real and what is not. Command line or computer/fantasy invention of a photo or video is NOT real and may be unethical on a case by case basis and will eventually implode if it is left to its own course. This is why. I call it the Red Duck postulate. Suppose you and 2 million friends did a command line invention of a Red Duck and all of you got a fantastic image of a red duck in return. Now suppose that the global compendium of duck images (say brown ducks) totaled 1 million duck images and in a matter of a few weeks your friends uploaded your two million "photos" of red ducks onto the net. AI training itself on ducks would begin to think red ducks are natural when they are only AI inventions. The model then begins to collapse on itself because its been fooled by 3 million ducks of which only 1 million are real. So the AI training model collapses into unreality or better termed an image idiocracy. Furthermore small children will be told by AI that red ducks are real when they are not as the line between reality and fantasy is blurred beyond hope of recovery causing a backlash by the "organics" i.e people that have seen a lot of ducks in nature but never seeing a red duck do not believe they exist despite what AI now tells people.
Eventually, the quality of AI will not be affecting by learning on self collected media vs public/paid content.
With the software getting better, it will need less content to learn from. Eventually, a big AI company could suffice with a tiny studio collecting exactly the kinds of source media to train its AI off.
Disagree? Please tell me so I can learn.
from what I understand, its a good software, but good equipment just cannot be undermined, the enhancement seems to be laggy which probably can be fixed but that's extra work, and when it comes to stabilization, you can't compare anything to a gimbal, though I like topaz over warp stabilizer, rather rent a gimbal or a better cam for a job and use this software to enhance it further if needed
And.... it looks a *little* bit better. Let's not forget that the best technology in the world still cannot compensate for poor film-making.
Same with audio production... something I've dabbled with my whole life... despite improved access to higher end tech... 90% of what I do is still crap.
i can never see the differennce no matter what settings i pick .even this guy who knows what he is doing i can barely see the differences in this video. im not sure its worth leaving your pc runnng at 100% for 5 hours just to very very marginally improve a 10 minute clip.
Next time compare side-by-side
great review, thanks
And now Iris V2 is such an improvement!
I'll be testing out Topaz Video AI 4 when it gets released!
4:44 Ouch my EARS BRO!
Difficult to take this seriously when the thumbnail is fake, having just blurred the original image for the "before" image, and is unsurprisingly not in the video. "WTF" indeed.
Its in the video actually 😂
So we're supposed to just ignore your floating head at 0:43
Ok then.
Nice video though.
how you only got 68 subs? this is really top notch quality
Thanks man! More vids coming soon
That "iris model 2 passes" looks super fakish and is totally unusable in my opinion.
On the slomo side there are no wishes open for me on davinci resolve, which is doing a awesome job.
I see what you mean, I think you can get better results than I did with some more messing with the settings, plus the model keeps improving. New Iris model just came out as well.
Original video, focus disappear in 3, 2, 1...
Did I just see dude's head detach?
Nice video, good work
Thank u for the video.. Very informative.
The Iris looks to smushy for me. It looks like an instagram filter, or Oil painted somehow.
Good review and good point about video "technique" being really computer "power abilities" but should we all consider how many current very succesful singers and music stars are just a result of autotune trickery? No one cares about that as they sound great, so why should we care about video lookiong perfect ? Isn't that what we want when not having big producer's resources ? Greetings from Paraguay.
People who throw ALL A.I into the same category of “stealing art” are ridiculous. Like please show me where the “stolen art” exists within he final result? It’s only referencing other videos to understand what hair detail skin detail and eye detail looks like. It’s like me asking someone to draw a man with blond hair and blue eyes then being upset that they are using memories of people with blond hair and blue eyes to know what blue eyes look like and what blond hair looks like even though the end result looks nothing like any of the references they have seen. Some A.I models are terrible and really just an image compositing algorithm where you say “Santa riding on a plane” so it finds an image of Santa and finds an image of a plane and does a quick and rough composite of both images. That’s NOT the A.I most people use most algorithms if you put the same prompt will find 3,000 images of Santa all by different artists then 3,000+ images of different planes and will recognize the similarities between all 3,000 images to understand what Santa looks like then it will rebuild a version of Santa based on that. Can’t say “BUT HE’S WEARING RED!” Yeah 99.9% of Santa’s are. Tell an artist to draw Santa and they’ll be doing the same thing.
I'm not making a living from this, so I welcome these advances.
Welcome to our fully-rendered plastic world, loosely inspired by reality
the more similar an AI works like a human, the more obvious should be that complaining humans complain about themself. this also means: the more the AI generates photos, videos and stuff the more the AI learns from its own generated stuff ...
First video first comment... just wanted to say.. u did an amazing job.. keep it up!!! May you grow to millions!!! Subscribed
Appreciate it man!!
nice, but i would rather take the shot again to fix focus ;-) And is less work.
when filming documentary verite there are many times that filming something over again is impossible. This has amazing implications for doc filmmakers that might have otherwise thrown out an entire part of their story due to a technical error which happens when in the field following subjects for hours on end.
Dude get an external Monitor
funny how Topaz Video AI didn't do shit for my videos I wanted to enhance, it actually made them worse. Other AI alternatives were even worse and apps crashed for no reason. Topaz destroyed the sound, So I had to recombine video and sound after. Not a good experience at all.
I think you had a bugy version, Topaz pass trough audio (muxing) without conversion, works very well. Video depends on source material, choosen AI model and parameters. For (cinema) movies, keeping max naturelness, use "GAIA" or "Theia - fine tune" model. There are others cases where Iris and Proteus are not a good choice. I think Topaz should focus here more what you can do when "reconstruction" of something fails, respectively a model that can handel better such sources.
what the hell is happening with your neck, homie
no, too slow
"It takes away the skill level."
TOUGH! Suck it up. The progress of technology and the world shouldn't be held back by a bunch of crying babies.
не в фокусе 😆😆 (в фокусе не модель , а задний план за спиной)
At SLOW MO, Davinci with WARP would kick Topaz and Premiere to corner and kill them.
If using AI isn't ethical then neither is digital and the only true form would be film.
scammer