dpullum Loc: Tampa Florida

jlg1000 said in part: Everybody will download the "wedding app", ... Maybe the married couple will pay $19.99 for that service."

Wedding attendees are annoying to wedding photographers. For one wedding I did, I suggested to those who arrived early that I would incorporate their photos into the wedding TV program which was what I provided to the couple. I pass out my card and yes I got a lot of good phone shots and camera shots from the attendees.

I did not supply prints to the couple but rather a CD slide show that they could play on their big-screen TV... they loved it.

Jig1000 is this what you meant:
https://apps.apple.com/us/app/wedding-photo-app-by-wedbox/id1180914615

riderxlx Loc: DFW area Texas

These are very nice images of a lovely young lady. For the life of me I don’t know why those very tall, high-heeled shoes are worn. I was told many years ago. They do that to pump up their butt cheeks to look more plump. I think the shoes just looks silly and extremely awkward to wear and walk around in. Anyway, these models are nice thank you for posting.
Bruce

jlg1000 Loc: Uruguay / South America

Short answer: of course.

Long answer:
1) Because of Gödel's Incompleteness Theorem and Heisenberg's Uncertainly Principle, it is absolutely impossible to create a perfect software nor model.

Even the one in your skull has very well known bugs. Optical illusions, poor judgement, confirmation bias, grabbing a can of tomatoes instead of one of beans... to name a few.

2) The best AI results are created combining multiple models. Th configuration I used for my "girls" starts by selecting one model that renders very well the body, and then detects using a semantic detector that flags details that are rendered poorly, like faces, feet, hands, knees, etc. Once detected, it re-render those areas with other models which are better suited.

I defined a dict of pointers that has this form: {body: model_1, face: model_2, feet, model_3,...}

Te bug consisted in the fact that de detector correctly flagged the knees, but incorrectly selected the *next* available model to the one needed.

jerryc41 Loc: Catskill Mts of NY

I heard the heels make the legs look longer. Whatever the reason, I found that I did "look longer."

DirtFarmer Loc: Escaped from the NYC area, back to MA

To the point: AI is not a standalone system that will produce masterpieces. It takes operator skill to get the best results. Expecting the hoi polloi to flood the market with great works is premature. IMHO, AI (as hyped) is not ready for prime time. I don't expect to see it in my lifetime (although I'm old, so that's not an extensive length of time). But I believe it will improve.

riderxlx Loc: DFW area Texas

Hahahaha, you gotta a foot/fetish Jerry ?
bruce

riderxlx Loc: DFW area Texas

Oh my !!!! are these AI generated images ????? if so, were created by using a REAL WOMAN ? I seem to be missing something here and just have not got it yet.
Lost in the fog Bruce.

jlg1000 Loc: Uruguay / South America

1) Yes, these are AI generated images, I did them
2) No, no real woman, nor camera was necessary
3) I do not create them individually... once the "configuration" is ready (and the software bugs are ironed out) I can let my server run amok and create one after the other. It takes about 3~5 minutes each rendering.

Here are other examples of what I call "Pose A" "Pose B" and "Pose C".

As you may notice, I still have some imperfections... it will take a bunch of programming hours to fix them. Nevermind, this is just a side project for fun.

Btw, anything can be created... look a the fourth example.

jlg1000 Loc: Uruguay / South America

Short answer:
Of course - just as any kind of art or craft - it needs operator skills...

Long answer:
... *BUT* here comes the interesting part: once the system is set up, I (nor any other AI operator, btw.) needs to do anything. The computer will render one image after another, all quite different, but based on the same "style".

Just think how long it took to Renoir or Van Gogh to develop their personal style and later create consistently masterpiece after masterpiece. It took me a couple of hours to create the "style" (configuration is a better word) that renders those women. I left the computer run overnight, and the next morning I had about *700 images* ~50% of them, usable.

Long Answer II:
Unless you are *really* old, do expect to see reaching AI "prime time" status. At the current development speed, it is expected that an AI image won't be distinguishable form a pro photographer created one in 12~18 MONTHS, quality short video clips in 18~24 months and full feature movies around 2030.

I am pretty confident of those predictions, because I'm one of the guys hired to make them.

Ava'sPapa Loc: Cheshire, Ct.

Nice job (very nice actually) except for the right foot on the left leg in both of these.

riderxlx Loc: DFW area Texas

Thank you for the reply and explanations. I thought this was AI generated by reading but now that I am looking at the images again this time on my desktop puter I can kinds see this. In the morning I review the daily forum in bed on my iPad so the sharpness is not quite as good as the 4K monitor. Damn these are hard to tell if real or AI.
My gosh, I still prefer real women though but I can see both good and bad uses for this new technology. For the good I hope but unfortunately the bad will find a way to use it for bad.
bruce

mwsilvers Loc: Central New Jersey

Photography is optical. Cheap lenses with poor optics enhanced by software can be made to look better. The use of AI may even fill in the gaps and make the output from poor quality lenses look very acceptable to most people. However, if software, and AI in particular, could make mediocre lenses look better, imagine the results if the same technology is used on top quality optics. It is unlikely that the desire for top quality optics will somehow disappear.

jlg1000 Loc: Uruguay / South America

Totally agree...

... just one caveat. Imagine that Gucci needs a shootout for their 2025 summer collection.

Supplier A, hires a bunch of models, photographers, make up specialist, hair stylists, flies to location with all the stuff, does the shootout, edits, etc. etc. I will probably take weeks and cost north of 100k (maybe *very far north*)

Supplier B is an AI software specialist, hires server time (~49.99 $/month) and renders thousands upon thousands of excellent high resolution images, aimed to every possible market (different races, locations, clothing, etc.). Now imagine that they he or she delivers overnight and charges ~10k.

Who will get the job?

riderxlx Loc: DFW area Texas

Well yea, money talks, BUT ! what is REAL and what actually represents the real product ?
Just a thought.
bruce

jlg1000 Loc: Uruguay / South America

A very profound question... I don't have the answer.

According to Heisenberg and Penrose, there is not even an answer...

This hallway is a new project... I'm still perfecting it.