We can't believe what we see anymore. The photograph's legendary credibility, which has been steadily eroding for the last century, has finally hit bottom with the introduction of generative AI. Images and videos can no longer be submitted as courtroom evidence. Seeing isn't believing. The value of an image has plummetted.
In thinking about the future of imaging, I started considering some of the forgotten benefits of film. You have an original negative which bolsters the image's credibility. B&W images, unlike today's color ones, can last 100 years or more if printed using museum-quality archival techniques on acid-free paper (a fact that has been pretty much forgotten by the masses.) And I grew up shooting film and working in a darkroom, so I'm more qualified than most to bring back this time-honored, old-world craft.So I've been toying with this idea of starting a new brand of portraiture business, one that goes BACK to the days of film and hand-made darkroom work and marketing it to the millionaires who spend their summer vacations on Cape Cod every year. (We live just minutes from Cape Cod, so it's a good fit.)
As a test, I threw this prototype website together to test the messaging. What do you think?
If I do this, it probably wouldn't be this year, for it would mean I wouldn't be able to travel during the summer. And the thought of going back to darkroom work really turns me off. (I spent WAY too much time in the darkroom in my youth!) But I believe this kind of an idea can really stand out in the modern era of smartphone images that carry relatively little intrinsic value. And it's a potential answer to all of this AI-generated crap.What do you think? Will this idea fly?
===
Oh, wait... Maybe Generative AI is the answer!
Ever since I wrote the first overview of Generative AI for Cameracraft Magazine back in 2023, I've been kind of obsessed with the ideas and started following the industry closely, waiting for the tools to reach a certain level of maturity. Some of my favorite experimental videos from other people include shorts that portray Elvis Presley in various pop-culture roles: As Matt Dillion from Gunsmoke, as Dr. Who, in the movie Jaws, auditioning as Han Solo in the original Star Wars movie, and as the Captain of the Enterprise. (Also, while I thought the AI-generated song I used in my Uganda post was impressive, that's nothing compared to this track of "Baby Shark" done in 1950's Soul style).
This is the future of filmmaking. One person and a keyboard. But how is it done?
| Made this last May. Almost there but not quite... |
And this is one of the bigest problems with generative AI - it's a very hit-and-miss proposition. For every great image you may see online, there were probably 100 failed attempts to get there. You're at the mercy of the algorithm, and you generally have little control over the fine details of the final image.
But there are tools that offer you more control than just typing a prompt. One of the most important ones is something called ComfyUI, a stable diffusion front-end that gives you lots of control, providing fields (called 'nodes') for both positive and negative prompts, specifying which models to use, ControlNet Nodes for image conditioning, Utility Nodes for smage manipulation, and various Custom Nodes for advanced features. It's so complex that it reminds me of the analog Moog synthesizers of the 1960's (in fact they're wired up similarly!). It's not easy but the process gives you unprecedented control.
| Learning the complexities of Generative AI is analagous to learning the Moog synthesizer in the 1960's. (Below.) But it sure gives you the kind of control you just can't get anywhere else! |
| Moog synthesizer from the 1960's. |
| Me at various stages of my life. |
| Our youngest grandson with his idol, Dwayne "The Rock" Johnson |
| The "Before" image: This was last year's attempt to create an image of Carol and her father (who died in 2010). The likenesses are awful and it looked like it was made with AI. |
| Here's the new version, created using my new methodology. It looks like a real picture instead of an AI image. These tools are finally ready for prime time! |
| A daughter reunited with her long-deceased father. |
| We had two cats over the years who have since died. They never met each other. Here's an image of Carol holding them both. The likenesses are amazing. |
| A better image of the King of Busoga |
| No AI that I know of can handle this many image uploads and still retain likenesses. |
- Parents with grown children who want a "snapshot" of how their kids have aged through the years.
- Parents who have lost children can have an up to date family portrait showing, which includes updating the visible age of the participants. (See my blog post about rememberance photography to learn about how valued such images are to the families.)
- Widows and Widowers can smile a little when they see themselves reunited with their spouses.
- Pet owners who wish to be "reunited" with their best friends.
- Large family reunion composits can be created.
- It can bring some comfort to families with military members who are POWs or MIAs.
No comments:
Post a Comment
Thank you for your comment! All comments must be approved by a moderator before they will appear.