Predictions about the impact of AI-generated imagery

[ahalpert]

Where do you guys think AI-generated imagery will take us? I have my own thoughts that would fill an essay, but what do you guys think?

My basic feeling is that anything that isn't essentially documentary or doesn't rely heavily on human expression and emotion will be open to being done with CGI instead of actual photography.

Just saw this about creating a digital human presenter, marketed as a way to save money on video production. Looks and sounds robotic, but it will get better. And it will certainly get good enough for cost-conscious clients who can live with a substandard product.

https://petapixel.com/2022/09/20/new...-single-image/

Here are links to a slew of recent articles I've read about the advances in AI-generated photography and, to a lesser extent, video:

AI landscapes:
https://petapixel.com/2022/08/16/the...ted-by-new-ai/

Find out if AI trained on your photos:
https://petapixel.com/2022/09/19/you...ge-generators/

AI can now edit human faces:
https://petapixel.com/2022/09/20/ai-...t-human-faces/

thousands of AI stock photos are for sale:
https://petapixel.com/2022/09/16/tho...hoto-websites/

Photography website bans AI-generated images from its platform:
https://petapixel.com/2022/09/15/pho...-its-platform/

Man asks AI to recreate his photos based on text prompts:
https://petapixel.com/2022/09/14/man...re-astounding/

AI images raise ethical and economic concerns:
https://www.axios.com/2022/09/12/ai-...able-diffusion

[DLD]

If one writes. "two visibly inebriated white men in their late 30's are having a friendlybut intense face-to-face conversation at an airport restaurant. One of them, Misha, is wearing a sports jacket; the other, Sasha, a wool winter sweater. The table itself is under a heavy weight of four vodka bottles and two 750 ml bottles of red wine" , then one should get something like this. (skipping the besom, the half-finished glasses of wine and some leftover on-table snacks)

hqdefault.jpg

then one should easily be able to alter the POV, i.e., the closeups/OTS vs.wide shots and the height and positioning of the camera.

And there's your movie.

[NorBro]

All of the above and similar will become the standard by default. You and I and DLD and others will move on in life while newer humans will be born into this kind of matrix and technology and it will simply be the way to do things.

You'll still see some 2022 vintage methods and practices, but there will be little desire for them when automated technology can achieve mostly the same purpose, the same end-goal.

And that AI stock is incredible. Who could possibly know the difference, care if you don't appreciate the art, the skilled labor?

Definitely not 2100 humans.

[Tom Roper]

I am already using AI generated text to speech human sounding voices to promote my upcoming web based wedding videography business for 2023. A logical next step would be the AI digital human presenter, and I think I will. One of the national wedding companies I shoot for already uses AI to produce wedding videos from my camera footage. It's not unlike what my iPhone does when it creates video memories, puts slides and videos to music, subtly zooms and moves. I can create intelligent wedding videos because of course, I have human intelligence. But the videos produced from my wedding shoots by the AI based systems, are intelligently produced also, and do a good job with combining appropriate music, timing, zooming and speed variables, slow motion, fast motion while being indifferent to frame sizes, frame rates, codecs. It's like the bulk audio and footage is dumped into a bin, then sync'd automatically, makes scene cuts. Doing this in Premier or Resolve will be hard to compete with this because AI can do in a few hours what could take me a week or longer.

[scorsesefan]

That youtube video gave me the chills

[NorBro]

As far as the audio, you must not watch TikTok...similarly that's how a lot is over there with the main text-to-speech VO from the Canadian woman.

It's so robotic but eerily natural and almost comforting on that platform.

[DLD]

I'd think that in terms of audio or generally defined "acting", one would be able to choose not only any type of voice - Brooklyn American, Russian immigrant, Spanglish, etc. - but every intonation as well. Anger, fear, sarcasm - you pick it and AI will deliver it. Likewise, if you want a digital Marylin Monroe, TM to star in your next film, you'd be able to hire her for a reasonable price, although Jayne Mansfield would do the same job for a lot less.

[NorBro]

You can...I've been using this for a few months. It's incredible although very specific editing is kind of a PITA.

Directing voice-acting is easier, quicker (telling humans what to do)...for now.

https://murf.ai/

[DLD]

after a little BS, the meat of the video starts about 3 minutes in.

[NorBro]

Most of the time I even like the subpar ones out there, so some of the ones I quickly heard above (didn't watch the entire video) are the real humans as far as I'm concerned.

[DLD]

I watched a whole bunch of AI generated voice clips ... and I am bored already. Maybe if someone has an accent - I have a slight one - and wants to sound local in whatever language, then text-to-voice is OK for an advertising type presentation. Otherwise, dialog has to be acted out. Maybe those Deep Audio Fakes will work but someone will really have to know his audio talent too. I.E, if you want to have a film starring Patrick Stewart, you'd have to write for Patrick Stewart. But it can be done.

[DLD]

And audio will obviously require less muscle than video.

[firehawk]

As this gets better I can see this replacing us to a significant degree. If it's affordable and fast, the 'old talking head corporate/commercial video could go this way. Why hire a crew to come in, set up, shoot and post when you can select your own CG actor/actress and do it the AI way.
Maybe we need to get on board and offer it as an option before the clients do it themselves? Kind of like making our own noose or digging our own grave though, isn't it?

Between the cheap kid with a camera and this AI to compete with, perhaps the only slice of the pie left will become the high end who want it done real, and really good.

[DLD]

As this gets better I can see this replacing us to a significant degree. If it's affordable and fast, the 'old talking head corporate/commercial video could go this way...

That would be a "step", but there are always going to be more after that. Computer talking heads are boring.

[NorBro]

Between the cheap kid with a camera and this AI to compete with, perhaps the only slice of the pie left will become the high end who want it done real, and really good.

I think so but that high-end won't be hiring Joe or Jane from DVXuser. They usually require small armies of people and equipment. I guess simpler high-end clients exist, but I see those disappearing more and more as the years go by because the people in charge who grew up with VCRs and newspapers won't be making the business decisions anymore.