The Truth About GenSwap - First Look How This AI Tool Changes Our Photos
ฝัง
- เผยแพร่เมื่อ 6 มิ.ย. 2024
- Change objects in your scene with AI! To access the new generative AI tools, including GenSwap and GenErase, get the Luminar Neo Creative Pass or Subscription with a discount here: bit.ly/GenErase If no other promo is on try ATNEO10.
GenSwap is Luminar Neo's AI image generation tool that allows us to select and remove objects from a scene based on a text based prompt we enter. This is my first look at this new cutting edge technology. Write "Prompt" in the comments if you'd like a more indepth follow up explaining the technology and the best methods for getting the results you want.
Adobe Photoshop has been rocking this new generative technology for a while so it's great to see Luminar punching back with the first of it's own generative tools. GenErase is designed to replace the area we paint over with the click of a button. As a comparison, check out Adobe Photoshop's new AI tools at the best price here: bit.ly/CC-adobe
👉 USEFUL PHOTOGRAPHY LINKS: 👈
𝗥𝗲𝗰𝗼𝗺𝗺𝗲𝗻𝗱𝗲𝗱 𝗦𝗼𝗳𝘁𝘄𝗮𝗿𝗲
🏋🏻♀️Get Lightroom and Photoshop here: bit.ly/CC-adobe
😄 Lightroom Alternative without subscription!: bit.ly/dxo-editing
💣 Insanely good AI sharpening, denoise, and upscaling: bit.ly/TopazPhotoAI
𝗣𝗹𝘂𝗴𝗶𝗻𝘀 𝗳𝗼𝗿 𝗽𝗵𝗼𝘁𝗼𝗴𝗿𝗮𝗽𝗵𝗲𝗿𝘀
🤩 Luminosity masking panel for Photoshop: bit.ly/3dpGeub
💪 High end finishing and effects bit.ly/nik-collection-dxo
𝗨𝘀𝗲𝗳𝘂𝗹 𝗟𝗶𝗻𝗸𝘀
✏️The graphics tablet I use amzn.to/3SXfVhd
⛅ Best Sky replacement pack: bit.ly/2Ruj8KP-Skies
🖱 Tourbox - a tool for editing quicker: bit.ly/39TdQ1S (discount code: TB20210408XXL08)
🖥️ My favourite colour accurate monitor: amzn.to/3UbFgVm
🎵 Music when featured is from bit.ly/ATHookSounds
These are affiliate links to the equipment I use - they are not sponsored by the manufacturer in any way. If you use these links it costs you no more than normal, but I may get a small commission which helps me to keep providing free content for you and my viewers. Thank you for supporting the channel and free education. I really appreciate it. 🙏
![ไม่ต้องมีที่ที่ให้ฉันอยู่ แต่ขอแค่มีฉันอยู่ก็พอ (Spaceless) - getsunova [OFFICIAL MV]](http://i.ytimg.com/vi/4N8idbxZ-7M/mqdefault.jpg)
BIG NEWS! Skylum has just launched a summer sale where you can save a massive 81% off Luminar Neo here bit.ly/NeoSale But the cool thing is in addition to that they've given me a coupon code to share with you so you get to save ANOTHER 20% (use code AFF-T5C1C0 ) I really hope that helps you out! It should work for both the perpetual licence and the subscription too. Enjoy😄 Best wishes, Anthony
The possibilities with GenSwap are truly fascinating! Loved how you explored different prompts and masks to transform images.
Cheers Anthony, best laugh i've had for a while! I am sure things will improve!
Haha. Good to hear! 😀
I will have a lot of fun with this,can,t wait
As you say it is in early stages and a lot of fun to be had. This opens up where does photography end and creative picture begin..
Looks like a real gimmick - I'll be steering clear of this one!!!! NEO is supposed to be a serious photo editor not a gimmicky AI cartoon like program. This would be more at home in Luminar AI than in NEO. I hope this is not the way they are heading with more silly stuff instead of improving the serious editing side
The current best deal for Luminar Neo can be found here bit.ly/Luminar-NEO. You can try code ATNEO10 for an additional discount.
Fairly glad I'm not paying a subscription for this tool, hopefully these tools will evolve further than the previous tools that seem to have been abandoned in a barely functional state (upscale for instance)
if luminar stays consistent there will be little to no improvement as their masking tools still fail mostly, powerline erasure remains non functional almost in all cases so here we have a bright shiny bell to boost new sales
@@daviddyephotography Funny you mention powerline removal. That was one of my biggest beefs: if the lines are sharp and distinct, the tool works well. But if they're out of focus (like the lines are closer to you than the subject), the tool fails miserably. And shiny bells and whistles are what every company uses to attract new sales.
I like/love the erase and expands on PS and will check Luminar out as well but for actual content generation, Stable Diffusion + ControlNet is light years ahead of the commercial guys...
I've done a few tests with PS vs Neo (which actually runs Stable Diffusion) and there's not too much in it in terms of result. I'll try and get a comparison video put together.
@@AnthonyTurnham Oh that really would be interesting. I have really been making a lot of use of the erase and canvas expand features and they have worked so, so well. But generating even the simplest thing, like, I dunno, a cola bottle standing on a table generates the most outrageous nonsense 😂 With perserverance and NOT throwing the baby away with the bathwater, it can do things, bit by bit, but a dedicated Lora in SD will usually do it faster.
@@AnthonyTurnhamA comparison video would be great!
Doesn't look too good to me. I think compositing would be better. Good job on the video though.
you should compare this to generative fill
Thanks for the video. My luminar neo doesn't appear Genswap! is it an extention?
Hi. It's in the library section. It's available with either the subscription, or with a creative pass if you have bought the Neo editor outright. I have a discount link/code in the vid description if you're interested.
Yes Show more of GenSwap because to tool didn't seem to be to functional as it is now
When is release? Can probably be useful in certain situations but I agree with some of the other comments here. Luminar need to spend some time improving some of the previous features before releasing new ones.
16th Nov is the release date.
Thanks for the preview. Since my subscription recently renewed I guess I will get his but based on your examples I can't see me using it except out of curiosity
Yeah, I'm still pondering the practical applications for it... Certainly a fun tool to play with though.
@@AnthonyTurnham Saw a brief bit of another stream where someone used it to put a reflective puddle on a city street. It looked a little more useful in that type of scenario.
Thanks. Not the most impressive or inspiring photo AI, but an entertaining video nonetheless. Would love an in-depth video on photo management /cataloguing for Neo / AI, particularly for someone looking to move from Lightroom Classic with a large catalogue. Cheers
as a photographer this TOTALLY useless, now if they had a tool to generate expanded boarder areas then maybe
Expanding an image in Photoshop allows you to do perspective corrections that are impossible on the original image.
I did a video last week and showed how gen erase swapped people instead of deleted them. Now they released gen swap? I'm confused why my generase is swapping too?
Hi Rob. One thing's for sure, the AI is easy to confuse. And at some point we may need to consider changing what the "I" of AI stands for 😆
@@AnthonyTurnham haha.. spot on.. I think they need to send in the clowns.. it will get better in time.. haha that clown photo was a crackup.. very funny 🤣
I had a recent photo of Vishnu at Ankor Wat, however there was an unwanted person in the background. I initially tried to get rid of it using layers combining the following photo (which also had an unwanted person in another location). But I didn't understand how to do this. Then! I remembered Generase yeh! But......instead of erasing my unwanted figure it replaced it with some other black figure which wouldn't have been out of place in a horror movie.😮
Yeah, the results can be unpredictable, comical and frustrating. GenSwap allows you to be more specific about what you want to replace the area you're removing.
😂😂😂😂😂
So I got to play around a bit with GenSwap last night and it was just unusable. Does Luminar have some kind of training about how to write effective prompts? I couldn't get it to do anything I was asking. I took a straight-on portrait of one of my kids, masked the eyes and ears, and prompted GenSwap for "sunglasses". Instead of adding sunglasses to the picture, it just ignored me and instead made my son's Western eyes look Asian. I did several other tests with and without prompts and I only managed to get the "AI" to do what I asked once out of maybe 30-40 tests. GenErase works really well, but GenSwap is just useless. I'd even say it's less than useless. Unless there's some trick to writing prompts that I'm missing, GenSwap will never get used by me. Maybe this is a situation where the AI improves the more it gets used. If that's the case, I'll give it another shot in a year or so when it's had a chance to learn basic English words like "sunglasses"...
Yeah. Good point. There is a lot to be said for effective prompt writing but at the moment I think its an issue with the fledgling technology rather than the prompt itself.
Put that scary clown in the background. 😮
Yeah, I'm a little concerned where my subconscious was when I was making this video! 😬😆
Cant see me bothering with this.
As a photographer for nearly 60 years I just don’t like AI in one’s own photos, it smacks of cheating and laziness. Fine for graphic design maybe. Then again maybe I’m just old fashioned, after all I still use film 😂.
It's a really thought provoking area around the ethics of using this kind of technology, how much is okay if any. It's certainly got me thinking. Personally I prefer the traditional methods too, where the final result is truly the result of your craft.
Huh! I only play with these tools when I'm on vacation, and (sadly) none of THAT until mid-January - but I WILL get a creative pass, and it'll be fun!
That said: I'm curious about what COULD be a very practical application: I work for the Reformed state church in Switzerland, a country with exacting laws about the rights of any person portrayed in a photo. For a church brochure we want to use a photo of about 20 children in a Christmas pageant - and it you look REALLY carefully, you can just identify the individual kids in the shot. We try to play by the rules, so we're currently in the process of getting usage permission from ALL the kids' parents. Result: We we could potentially get an OK from 19 parents, but simply not hear back from one family, and have to scrap the shot entirely. But: What if we used this tool to replace that one face? Seems like we'd be in the clear, no??
Funny you mention that scenario - I've had to do the same thing just yesterday to make an image usable from a legal point of view. I actually swapped the face and then swapped the lower half of the face again - just in case AI had pulled an existing person, so at least I knew my new "person" was a complete fabrication.
@@AnthonyTurnham Good idea to make it a composite!
BTW, when I mentioned the idea at work, people thought it was icky, and they were positive it would look horrible and cause problems.
On one hand, they're definitely wrong about how the picture would come across. Given a couple passes, no normal viewer is going to come even close to noticing a face swap if they didn't have background information about the shot.
OTOH, if you were said kid's parents and had maybe explicitly said no to usage rights - and then saw that we'd used the shot ANYWAY - but THEN noticed that your kid had been transmogrified into somebody else ...
There's no denying that this is a weird rabbit hole.
@@rather-reverend Yes for sure! I appreciate your predicament. Perhaps a full body and face replacement and then you're out of the firing line from those parents as no part of their kid would be present in the shot. But yeah, tough one!
try comparing with photoshop
Yes, I've been doing that and it's actually very interesting. They're both running the diffusion model so the errors that crop up in one are also prone to appear in the other too.
They should add some useful luminosity masks or improve object selections rather than playing with these fun but poorly useful tools..
I've been petitioning for luminosity masks to the developers but unfortunately it's not in their plans 😒
@@AnthonyTurnham I really liked the original intent, but now they are trying to emulate Adobe commercial strategy, but they are not competitive as Adobe in doing so. Really hope they will try to re-focus on photography, but it seems not. AI masks are just a joke for kids and they still aren't capable to optimize the resource usage...
I tried Gen Swap and was very disappointed. Generative AI is painful, so it's useless except for playing games. I think Skylum has a lot of work to do to make it a serious thing.
Hi. Agreed. I was playing a lot with Adobe's gen AI recently and that too suffers from similar issues.
It is absolutely terrible currently.
It is not exciting times in the photo editing world
Personally I don't love the idea of augmenting my work with AI creations, however I do think that breaking from the status quo and exploring new possibilities is exciting.
Wow, those examples are just comically bad. Other than adding something simple like the moon in your first example, this appears to be mostly useless. I can't see anyone spending any time with this tool beyond a few quick "just playing around" trials. The struggles you went through just to get something that slightly resembles actual human faces in that birthday photo will turn professional and amateur photographers off immediately. It would be one thing if the AI was generating normal human faces every time, even if you wanted a different look. But the "faces" it was generating weren't even in the realm of workable. No one is going to want to spend 30 minutes with a single photo tweaking and tweaking and tweaking and tweaking their AI prompt, hoping that something workable comes out the other end. I mean, how many times did you have to ask it to put a freaking clown into the photo before it actually generated a clown?! That's not even AI; that's artificial stupidity, like it was willfully ignoring the one thing you actually wanted. Maybe it works better with non-human subjects. Adding buildings, or cars, or planes might be more productive. As it stands, this AI tool needs A LOT more training on human faces. I'll be forever haunted by the image of a young boy with huge pink chicken lips now. Thanks for that, Anthony...
In all seriousness, though, I feel that this perfectly illustrates one of the biggest shortcomings with current AI. The prompts. People will need college degrees on how to write effective AI prompts. Just swapping one word, or changing the order of the words you're using, can have massive consequences for the outcomes. Hell, even using the exact same prompt as someone else will generate completely different results. The first company that can really crack the prompt interface to make it easy for the average user and produce excellent results nearly every time will be the big winner when it comes to AI.
Hi Matthew. Thanks for your thoughtful comment. Yeah, I totally agree!
Photography, really ??? 😮😫
I think if we avoid human faces, it should a bit better
With all this AI stuff coming out pretty soon you landscape photographers won’t have to brave rising at early hours and going out in bad weather to get your shot.
Just tell AI what scene you want and presto!
Sad, very sad if that’s where photography is going.
If serious photographers refused to use this stuff the app makers would stop putting it out, maybe concentrate on improving the other aspects of their software.
As for me, I could care less about some AI entity creating a pic a mermaid riding a horse. 😂😂😂