Beyond Colorization: modern tech to give you more realistic detail

Computer colorization.
Technical wizardry.
Painting a mustache on the Mona Lisa.
Burn the witch!

We're not only surpassing the clunky attempts at Ted-Turner-era to make older film more palletable to younger audiences, but there is a lot more detail and realistic processes that can give you a more (admittedly-artificial) "realistic" experience from earlier works from the crafts of film, art and even sound.

This is a great video that first explains all the processes the technician is applying to his samples: stabilization, colorization, frame damage and rate recovery, image enhancement, foley...a whole portfolio of techniques. And yes, all are not perfect, and he admits that...but the final results are promising:

 

Let's go further...back...and with the same concept...


Is it realistic? Well, no - NOT YET. But, is it a closer look at something we have never seen in our lifetime? Well, DUH - closer than we've ever gotten to the original.

 

You can even apply AI restoration to statues and artifacts:

Now we are seeing what an application of earlier technologies can do, when taken into a new extreme.

As you can tell, I've been a fan of these resurrection crafts and the results shared online, for some time now.

 

I saw this done with Roman emperors. Very spooky stuff.

 

Let's take another resurrection craft, to bring the human voices of the past to life, from a forensic expert:

 

And the more obvious recovery technology for this board:


Real historical artifacts considered unplayable for over a century.

 

"Nor do we really realize their true potential for the first time..."
Images capturing sound waves on paper, "played" by scanning into audio software:

 

And naturally...all this fascination with using modern technology to recover more details from what exists already recorded in our past...comes from my amusement at every huffy film "connoisseur" who ever pooh-pooh'd at Ted Turner's attempts to simply recover more revenue from a film library he actually had ownership of...and never destroyed a frame of the existing original, in order to do so.

This is the same amusement I get when I see people wetting their pants here every time Giles Martin releases a new version of something that already exists in their own collection in multiple form...and then starts screaming the sky is falling, as if their own originals are going to somehow vaporize themselves in a puff of entropy, due to the existence of a new edition coming onto the marketplace.

We need to take the development of recovery and enhancement technology seriously, and not jump to grabbing for the pitchforks and torches to storm the Bastille every time somebody tries to bring us a closer look at things we did not have in our possession before. We still have those possessions...and no amount of 4K Blu-ray, DES experiments with mono hit singles, or Steven Wilson remaster surround projects, are going to take those away from you.

And now if you'll excuse me, my new all-in-one printer is on its' way over...and I'm going to scan some century-old family photographs, and clean them up in Photoshop...so, suck it.

 

Alright! Let's FINALLY get Citizen Kane, Casablanca and A Hard Days Night the high-tech full living color treatments they deserve!



Sorry, couldn't resist.

 

Annnnd, here's something in serious need of a remastering...Steve...?


I'm thinkin', the usual Hoffman-midrange-goose trick isn't gonna save this one...

 

It could be better judged if it was applied to paintings of people who've lived in the age of photography - the present Queen, maybe, or Winston Churchill.

 

M'kayyyyyy...


Why would you need an AI recreation of a painted figure if you already had a historical photograph to source from? I don't think you're getting what it is I'm talking about here...I'm talking about getting a better impression of the subjects we have available, not about comparing quality control.

Maybe you have YouTube on your own computer, and could cherry-pick examples for yourself, so you don't start the, "yeah, but...what does it do with Presidents with moles on the left hand side of their face..." gotcha game.

 

The modern representations look very little like the sculptures. They don't look like the same ethnicity, for one thing. They look too smoothed out and bland, and some features don't even match (nose sizes/shapes in some instances).

 

I’m with Slackhurst here. Whenever I see a process that’s supposed to reveal new information about old things, I want to see it run on a control just so I have a clue of how good it actually is at what it’s supposed to be doing.

If the painting-to-photo process is run on a well-made painting of, say, JFK, and the result looks at least as much like JFK as the painting does, then I can see that the process actually works. But if the result ends up looking less like JFK than the painting did, then I’d know it’s a busted process.

 

Then by all means...go find a video that does exactly what you want to compare it with, and post it.

I'm not looking for comparison shopping here. I'm just pointing out that the technology has developed greatly since the days when people whined because It's A Wonderful Life looked bad on WTBS in color.

This isn't even a thread about AI and turning scans into cartoons. It's about the vast technologies that are giving us a deeper look into legacy media by pulling not only more detail out of what it wasn't designed to capture in the first place, but also approximates what we might be able to see, were it a more modern sample originally derived from more modern processes. It's not exact and authentic, nor do any of these processes even claim to be.

This is why I also included ancient sound samples, as well as both still and moving images. Movement that can look more like video than film. Images that are more stable and clearer than the originals. Audio the people who captured it in the first place, were not even able to hear reproduced.

You're not ever going to see what Sacajawea looked like when she walked, or what Thomas Jefferson's voice was like; that's a given. But, these people who use the modern tech to give us a better look into the past, can give you a better idea of what we missed out on.

 

Maybe they are giving you a better idea… but maybe they’re totally misleading. If there’s no control offered, how can you know?

 

What are the chances the craftsman is purposefully deluding you, versus a work that is doing its' best to create an image based on the information provided?

 

Here's another fascinating example of information retrieval based on the expectation that the originating document was intended to mimic what was measured. In fact...over an hour of examples.

Patrick Feaster "wrote the book" on images that have been recorded throughout history, from which sounds can be derived in methods the creators never dreamed. Most of these are sounds, just about all of them before Edison.

Some of them are simple animating of diagrams and graphics created to mimic what the observer in centuries past, extrapolated from technologies such as the primitive praxinoscope process, but applied to earlier drawings and other items.

And then, there is the "averaging" of faces in yearbooks, stretched over spans of two years, even four years. None of these are precise and accurate, yet still bring out a basic truth not apparent through standard observation.

I guarantee you, you have never seen such a detailed research on any of these in any of your regular studies of early sound...or the other media he branches out to. But, that does not invalidate the veracity of his results.

 

Here's a stunner. A television recording made in 1927...which was unable to even be played until a Scottish computer whiz decoded it into its' original 30-line picture in 1999.
This will take you 1:30 to get to the meat of the tape report:



Obviously, the mechanical disc that "recorded" the image, was not intended to be played by the computer that took the signal, which was inscribed into the grooves of the record.

 

Fascinating thread, Dillydipper.

 

It's because it's not so much about "colorization", as it is the complete concept of looking at media of the past, and finding new ways of getting more information out of it...even if the information isn't real (or, "accurate").

Most people who started reading, thought it was just about one thing, and they missed the thing it was really about.

In fact...Steve Hoffman made a serious career, out of doing exactly the same thing.

 

I totally "get" it and appreciate the historicity.

 

I don't care. A meme is not truth, nor is it a reasoned argument. It's just an out-of-context expression of something irrelevant. Save it for your closeup for C.B...

What exactly are you trying to say...and why are you trying to not say it with your own words?