Artificial Art Is Looking Too Real!
Вставка
- Опубліковано 30 вер 2024
- In this Critique the Community, we look at your best images created with Artificial Intelligence art creators. To view every image submitted in this contest, check out the full entries here: fstoppers.com/...
To enter the next CTC, head over to HTTP://www.Fstoppers.c...
Gonna start doing a dalle 2 vs mid journey on my insta stories to see how the prompts fare in both.
Really great show, gentlemen! Nice lighting & sound in the new studio; good flow, great conversation!
when he mentions its not up to par yet about the words he picks dont create what he wants. Just wait. its only going to get better and A I will get more intelligent and better quality over time. We are definitely under estimating this and will be no longer.
It also blows my mind how unaware our colleagues are, back in 2017 when I first saw "this person doesn't exist" I knew the countdown started. It was such a weird feeling. I was very successfully producing stock for 10 years, but the end is coming fast and gladly stopped a two years ago. Even the big stock agencies I talked to had no idea what was coming and how fast. Thx for ringing the alarm bells even though no one wants to hear them.
Yeah, it's all moving super fast now too. Like on Monday, Stable Diffusion released their model for free, so people can run it themselves on their own computer (assuming you have a decent graphics card). And MidJourney tested a beta renderer for around 24 hours (Mon into Tue) that is more more detailed and cohesive than their normal one (though they said they want to keep the ability for abstract images, even if it means having two renderers developed in parallel). If the FStoppers contest ended even a week later, the top images might have been totally different due to the MJ beta.
I've already seen at least one primitive text to video AI out in the wild, I think GPT-4 for text AI is due out soon, etc. It's getting really crazy really fast.
These are surreal "paintings" that's why is looking so distinctive. Not sure what dataset they used but it should be a collection of really great drawings/paintings
MidJourney also tends toward being more abstract and painter-ish and even tends toward certain color schemes. DALL-E 2 and Stable Diffusion tend to be a bit more literal unless you specifically tell it not to be. MJ actually tested out a new beta renderer the other day which was more literal and honestly even better than DALL-E 2 in many cases. They said during the announcement that they still want to allow more abstract interpretations, and if they can't manage to incorporate both in one model, they'll keep developing both paths at the same time.
If that temp beta release (already taken down again for now) was available during the FStoppers contest, you would have seen stuff that was a bit more detailed and cohesive in the entries.
ARE YOU NO LONGER DOING CRITIQUE THE COMMUNITY EPISODES?
I have suggestion - since alot of your videos in the past focused on "high end" commercial photography, allways with the silky smooth hdr edits and pretty models, how about going the other way, and asking for pictures taken of things that wouldnt normally be considered "beautiful" or "pretty" and where the resulting picture may not be "pretty" either but intruiging in some other way - sort of how we moved away from classical pretty music to stravinsky and schostakovitch focusing on more strange, uncomfortable and complex emotions. - I feel like that has been a bit of a blind spot in the online photography world, because everyone is going for either pretty and beautiful or pretentiously edgy with a contrived, nonsensical set. - There probably is a lot of authentic weird pictures, that people are just never asked to share.
- would be interesting and something new...
Great to see that you are keeping up to date with modern technology.
Crazy to think in the new hipster thing to buy and own in the not too distant future will be none AI created art and photography 😂
I dunno, all this AI "art" looks like shit they would sell in Target. Also "Cola" is the Spanish word for "tail" as in an animal tail. Marijuana buds are commonly known as "foxtails". Colitas is a plural diminutive form - so the line means the smell of buds.
A lot of the value of art is more about the interest in the artist themselves. So while AI art may be good enough for some purposes, it will likely not have the same value without the human factor behind it. The human story is often a significant part of the value.
Yeah, though I do think as time goes on, the AIs will allow more fine-grained control. I'd guess you'll be able to be super detailed and even have a conversation with the AI "move the cup over to the left an inch.. the paint strokes should be slightly thicker.. add a touch more red to her hair.. position her left hand mid-thigh instead of hip level", etc. At that point, you can argue the human becomes a much more active participant, guiding the image to become the vision in their head. Similar to how a director can be considered an artist with a very specific vision, even if it is technically a collaboration of hundreds or even thousands of people.
And now, just a few weeks later, Midjourney has a new test algorithm that produces near photorealistic results that's essentially rendered the old algorithm obsolete. It's absurd how fast it's evolving.
Underwhelming. Not saying it doesn’t have uses or a future but not excited yet.
Keep an eye on it, since it is moving really fast. If you search for Olivio Sarikas's video on MidJourney Beta, can see some new images from an engine they were testing a few days ago, with better quality than MJ has had in the past. Considering how fast it has moved in the last six months, who knows where these will be in a year.
It is pretty cool, still I think it won't have a big impact on many commercial photos, as clients want to have their property or product shown as it is. And a computer can't generate their specific product. Also people want to have their portrait taken and not just any random person.
So you are not aware it is already part of Photoshop??? (Neural Filters/Latent Visions)…
It hasn't been totally released yet on PS. I believe you still have to be put on a wait list for the roll out. -P
@@fstopperslive6318 09:15
@@fstopperslive6318 also, use front facing camera sensor instead lidar…muuuuuuch better results
whoop there it is SHALL LIVE FOREVER!!!!!! LOL
Colitas might mean "colita de pavo" turkey's ass, it's a thing.. we eat tortas and burritos of colitas...
backstreetboys for life
I remember when National Geographic first put up a digital photo and when they first had a touched up photo. So when will we see national geo using a ai photo on their front cover??? are photographers, and cameras becoming extinct ???
Not completely. Photography replaced portrait painting but painting is still around.
Ai will be not able to replace event photography as a job for obvious reasons
With AI Art is everything an original? I mean yall talk about all the people who go in the desert canyon where they setup the photo the light at the right time and the guy hides out of view to throw the sand up in the air to sparkle in the light. And everyone is snapping the same exact photo there. Anyone can take that photo and there are so many of the same photo from the same place. But AI Photo.....thats always original isn't it?
This is true yes. -P
Yeah, though it depends of course. Technically all the photos in the example are different too, even if the light is coming from a slight degree different, or a rock moved an inch due to wind. If you ask one of the AIs to paint The Starry Night, they'll usually look similar to the original painting but details will be different like someone trying to paint it from memory.
I would rather pay a human. I have played with ai but a computer never properly understands what i want