1966 Sports Illustrated article on Jim Jacobs

So doesn't it seems like Cus and Jacobs were watching the same, if not worse versions of the footage than what we have today?

If Jacobs was making copies, wouldn't that minimize quality deterioration?
And today we have upscaling and post effects for enhancements.

Either way, the best quality of Fitz-Corbett is yet to come.
We will see it in extraordinary detail for many angles before 2050.

 

This is incorrect. Jacobs and Cayton didnt film those fights. Those fights were filmed by other parties for a variety of reasons i.e. to be shown in movie theaters, to be chopped up and clips sent out to be shown in the news or newsreels, etc. So they would still have to show where they purchased, refiled, and retained valid copyrights from the original owner whether that was bigfights, tel-ra, winnick, etc.

 

I dont follow. You seem to be confusing film quality with digital quality which is kind of like comparing apples and oranges.

 

Jacobs was being arrogant. Just because a guy is 93, it doesn't mean his memory isn't razor sharp. The film quality was crap in the early 1900's. I would trust the guy who was actually there in person.

 

I'm assuming that the original copies were more prone to deterioration because of their format.
So Jacobs made copies which helped preserve the quality. Jacobs and Cus were probably not watching the original, they were watching a copy made, and that copy is probably not much different from what we've been looking at for the last 10-20 years in documentaries and on the internet.

On top of that, we've actually been able to enhance the quality in recent years using post effects.
And it helps that these post tools are essentially democratized, because if Setanta makes a quality documentary about Corbett, but doesn't have the foresight to stabilize the footage, then some guy on Youtube can.

Do you think Cus and Jim were watching stabilized footage of Corbett and Fitz?
I very much doubt it.

Do have any idea as to the process in which Jim made copies?
As in which format the copies were made in? We could look at how the specific material used holds up to deterioration over time.

 

Imagine I took the 20 great singers and brought them into a room.
And then I played a recording on iTunes of Adele singing through an antique 1910 Gramophone.

And when the 20 singers finally say "What the heck is this crap?"
I would make the ultimate reveal: "Ladies and Gentleman, you're listening to the great Adele."

Astonishment. Gasps.

 

Nobody is improving the old footage. Stabilizing it and correcting speed doesnt improve it. It merely gets it closer to how it was originally projected. When those films were originally shown they were projected at the correct speed and stable. They are only unstable today because of sprocket damage which wouldnt have been an issue when they were newly printed. When you project a film you can control the speed so that wouldnt have been an issue for jacobs. The sprocket damage effects a film transfer only depending on how it was transferred. If it was step printed optically sprocket it shouldnt have jumped around. This was expensive though and i doubt he had these films transferred regularly that way. Image quality isnt going to improve over time now or then without a frame by frame digital restoration and thats not happening. Thats years worth of work. The good quality images you are seeing is because they are coming from first or second generation professional transfers not the amateurish transfers that have circulated for yrs among collectors.

 

Sounds like he was a miserable ***** of a man, not unlike a very select few of our classic posters.

 

Well if we don't have access to the version that was originally projected, and you are improving it beyond what is available today, than you are improving it.
Setanta is the only network to broadcast Corbett footage touched up with upscaled HD. Yet it wasn't stabilized. So stabilizing that particular footage is an improvement over other versions out there. It sounds messy perhaps from the perspective of a purist collector. But the end product is what it is. Stabilized upscaled HD touched up footage of Corbett-Fitz, some of the best quality you can find online.

Another example is your footage sample.
If you speed up the slow motion portion of the clip, it looks much better than the real time version.
Here's the difference:

The action in real time from your original:
https://streamable.com/0q6x1

The action in 'real time' using my slow mo version and cheating the speed and other minor touch ups:
https://streamable.com/c9cty

But nobody else had done that.

Technically your version is clearer because the state of the raw footage, but the post technique I used ends up showing the action in much better quality.

And you can certainly go beyond the quality of the original projection. Colorization is one example. But software in the next few decades will be able to fill in many of the blanks, and apply incredible enhancements to existing footage beyond what you can imagine. We will be able to see portions of Corbett-Fitz as clearly as we can see Klitschko-Fury today. There is already very early work of researchers using AI to turn old photos into moving images. With Moore's law, and the exponential increase in computing power, it's only a matter of time in this century or the next when we can relive some of these fights from ringside.

Here is some very crude first step work going on at MIT with this:

This content is protected

But I digress.

It seems like from what you're telling me about the transfer process, they probably weren't watching better quality versions of the fight than we are. Or do you think they were?

 

I dont pretend to know what quality Jacobs was watching but a good guess would be that it was a first generation 16mm film transfer and you arent going to get a better quality than that from youtube, setanta, or whatever.

You cant HD upscale anything. Thats a phallacy. Ive tried to explain this to you before. If I take an SD source and then convert to an HD source you arent improving the quality of the original source, you are simply bloating the file. The only way to get an HD source for these old films is to transfer the original in HD as opposed to SD. When i have my films transferred I get both an HD source and an SD source and because of the technology used back then, the lack of color, the fact that these are often copied prints, not a master, the naked eye really cant see much difference when projected on even a 70 inch home theater screen.

In regards to the speed of my film you are taking one very short clip to illustrate your point which is giving you an inaccurate impression. This because Corbett-Fitzsimmons was shot using handcranked cameras befor the advent of the cam installed in the crank mechanism that gave filming a uniform speed. So throughout Corbett-Fitz, the action speeds up and slows down by varying degrees over varying periods of time meaning there is literally no way to get a uniform correct speed without some variance. I did the best i could but the problem you run into is that for 10 seconds the film might run at 15fps then over the next 15 sevonds it gradually ramps up to 20 fps, then over 10 seconds it gradually slows to 12 fps, and on and on. The fact that these changes in speed are often slow and gradual mean that beyond the fact that you can apply one uniform speed change to the entire film (because you might get the majority of the film slowed down but still have other sections that youve now slowed down too much or not enough) but you also cant take individual 10 or 20 second clips that gradually increase in speed and accurately, grade that speed back down, you can get close but not perfect. So what you have is a single section that ran slightly fast or slow or whatever, or maybe even gradually increased or decreased in speed when I either squashed or stretched the speed in a uniform manner, but thats not indicative of the overall film. You would run into the same issues if you were attempting to adjust the speed of the entire film even if you were doing it in cross sections becsuse there is no way to accurately gauge the variations in speed as the camera man got tired, excited, etc.

Im not sure how you can take a project I worked on and then slow down a slow motion portion, which is slow motion for effect, not because thats how the film plays in the complete version, and then claim nobody else had ever done that. You took some highlights you found on youtube and played around with the speeds to form clips highlighting those guys. You didnt reinvent the wheel. Everything, and I mean everything, youve done in those clips has been done before only its been applied to the entirety of the footage that is extent of that fight. And thats in addition to taking all of the disparate pieces found all over the world and editing them back into their original chronological order.

 

I disagree entirely on colorization. Colorization adds nothing to the film because who sets the standards? Its all fantasy. Even if you have minor information like Jim Jeffries wore blue trunks against Jack Johnson, you dont know what color blue, what degree, you dont know what their skin tone was, hair color, the color of the clothes, what the color of the ring ropes etc. A perfect example is Mendoza's goofy adherence to the idea that Tommy Burns was jaundiced against Johnson. So now do you add a hint of yellow to his palor because some moron on the internet thinks so? How much? How little? Do you make him look like he has kidney failure to satisfy Mendoza's belief that Burns was handicapped in that fight? So who is making these decisions and why? Why not just leave the films alone in their original, unaltered state, and be happy with what you have. I totally disagree that this improves upon the original image.

I wont even go into the idea of taking a still photo of a fight and suddenly being able to turn it into an actual moving image that is a replication of the bout as it happened. Im not interested in either science fiction as it applies to boxing, or being able to watch Harry Greb and Gene Tunney fight a boxing match that a computer THINKS they fought. All that matters to me is what actually happened in the ring. A computer will never NEVER EVER be able to look at a photo or even 100 photos and accurately predict how Harry Greb beat Gene Tunney in his own style over the course of 45 minutes. You might as well get excited about them inventing a time machine because thats what you are talking about. Most of those old fights you might be able to find 10 photos of. A fifteen round fight encompasses 10s of 1000s of motion picture photographs. Whatever information a computer could gather from a series of 10 photos and then run through an algorithm to predict how a fight looked would be completely make believe. It wouldnt be any better, and probably worse, than what you could imagine if you close your eyes and concentrate real hard.

 

False.
Upscaling properly from 480 to 1080 does enhance the footage.

Example:

This content is protected

Upscaling isn't just stretching the resolution.
It's algorithmic, and you need certain devices to do it right. It "reads" the picture and produces pixels to "fill in the blanks." It also adds sharpening to the image. And the end result is a better quality video.

I did the opposite. I sped up the slow motion portion of your clip to make it look real time, because that clip was more detailed. I don't know why it's more detailed. Was a second camera used?
In the real time clip, Fitz and Corbett look extremely bright, and harder to distinguish from the rest of the frame. The slow motion version however, looks much clearer in color and detail. Perhaps for silly reasons in how you edited it or something, but that's how it ends up.

Maybe someone else did all these enhancements in some production warehouse in a dimly lit room.
And I wish I could go back 100 years to see the quality of this fight in its original format.

But as of today, if "the rest of us" want to see Corbett in the best quality possible, I'm confident that my video provides that experience. I also stabilized your footage, and it improved it.

If all this footage were already enhanced, I wouldn't have needed to do any of that.
From a purist perspective, simply stabilizing footage using post isn't "improving" the footage from a technical aspect.
But when you actually watch the footage, it looks way better.

 

We don't consume data from media robotically. We experience it as humans.
Colorization adds a level of realism and connection that black and white simply cannot.

You cannot guess the exact color of objects, and liberties have to be taken.
But the viewer naturally assumes this for anything that is manually colorized.
They don't think "This is exactly what it looked like."
They think "This is what it could've looked like."

The notion that because we don't know the exact shade of blue in the sky, that colorizing it is a pointless endeavor, is a silly irrationality.

Despite these colors not being exact, people still love videos like this:

This content is protected

The colorization adds tremendous value and creates a rich viewing experience.
Look at the comments on the video. They're not talking about how the colors were picked.
They're talking about what it must've been like to live there.

I don't think colorization is a replacement for original black and white footage. But rather an augmentation.

Then you don't understand big data and AI well enough
It will never be 100% (Not in a way I can think of at least)
But they can get to 99% probability of accuracy.

And you won't be able to accurately reinvent entire fights (Again, not in a way I can think of.)
But you will be able to see portions of fights based on photographs.
And perhaps ultra realistic recreations of the fight, where you don't see what happened, but rather what it could've been like.

In the early stages of this, you will need better data to pull it off, so Greb will be tough as there is such little footage of him.
But eventually, the data will get so rich, and the AI so smart at predicting, it will be able to show us Greb fighting to a very high degree of accuracy.

If the data set includes trillions of examples of human movement, billions specifically boxing related, it will be able to guess very accurately Harry Grebs next move.

Things will probably get "tagged" with some measurement representing the probability of historical accuracy.

 

No, you are wrong. Bloating an SD file and adding filters like edge enhancements etc does not improve the quality of an SD image. I'm sorry but that's just a fact. It might make it easier for someone like you to watch on a larger screen but it if you did a side by side freeze frame comparison of the same image SD, and upscaled you would see there is no greater clarity, depth, or sharpeness to the image and that's what we talk about when we talk about HD. The "algorithims" you are talking about are nothing more than filters designed specifically to make the image "pop" a little bit, to trick the eye. That's it.

Regardless, the point is you didn't do anything new or different. Its not more detailed and it wasn't filmed in slow motion. Its literally the exact same film slow or fast.

Whatever, if you want to believe that somehow you took a low quality youtube clip (and I refer you to when I originally posted that clip and stated quite clearly that youtube wouldn't accept it in its highest resolution) and somehow upscaled it and improved it over the original so be it. Its funny really because even the version that you seem to think is HD is badly pixelated.

You crack me up. You only "need" to this in your head because YOU don't have good quality footage of the fights you are interested in because they have been circulating around the trader network for years in crappy multigenerational copy versions. Do you think these fights that exist on film to be transferred and ultimately end up on youtube exist all pixelated and sped up etc? They don't. What footage exactly have you "enhanced"? You are literally taking the best quality clips you can find off of the internet and tinkering with them. You really think the best quality clips you can find on YouTube are indicative of the best quality these films can be found in? Seriously? Here is a perfect example of what you think is an improvement:

Below is the original 1926 film commercial film print of Dempsey-Tunney 2 that I had transferred in SD:
https://flic.kr/p/U7KPjN
Same film transferred in HD:
https://flic.kr/p/UbjJa4
A screen grab of your 1080p "enhancement" of the best quality version you could pull off the internet:
https://flic.kr/p/UbjJ7D

The only thing I did to those was to crop them in order to fit the dimensions of your video. Blow those up all you want and you will see a significant difference in clarity and quality. Now, do you really want to argue that you did something earth shattering here? And keep in mind that the 16mm commercial print has only half the clarity and quality of the 35mm theatrical print which still exists and is quite common. So even the best quality that I have as an example, which isn't the best quality available, is noticeably better than your "enhanced" version.

But by artificially adding color you can alter a persons perspective of events incorrectly, as illustrated. Ultimately these films should serve as an impartial witness. By adding color you can change that. If Mendoza had his say Burns would be yellow fighting Johnson. Johnson would have bruises and swelling all over him against Ketchel. You can alter a persons perception dramatically by what you are talking about and I think that's wrong. To pretend that criticizing this is irrational is nothing more than a self serving argument. The same discussion was being had, likely before you were born, when they started colorizing classics like Its A Wonderful Life, and I think history will show that the majority felt it was a disservice to the original films. However, those are works of fiction whose journalistic importance is minimal compared to a historical film of a non fiction subject and as such I think its irresponsible to pretend that altering the films merely for your personal enjoyment is anything but reckless in this day and age.

Bull****. You are living in a fantasy world and the problem with this is that there is no way to check this kind of "accuracy" we have nothing to compare it against in the case of guys like Greb, Flowers, Rosenbloom, Sullivan, etc.

Nonsense. Come see me in 100 years and we'll talk.

 

Colorization, if done right, does add to our impressions from the film, even if you guessed the color of some clothes wrong or something.. And a lot of it can be done right by the right people, by using our knowledge about real-life objects (which have the same color as they did a century ago), by looking up historical artifacts in museums, etc. RGB/CMYK conversion to Grayscale is not a rocket-science either, there're books on color spaces by Dan Margulis, etc.

I also agree that smart upscaling of a lower-resolution video and brightness/contrast/color adjustment usually gives better results than watching the original, again, because of the way our brains interpret what the eyes see in motion. Again, this has been covered in cognitive psychology, etc, as well as in the same books by Margulis.

I don't think that neural networks have that much potential, though, although I'm not a specialist in AI algorithms, but common sense tells me it's impossible to do what you are expecting from it if we have just several static photographs or a video of training footage.