Blu-ray Review I Love Lucy: The Complete Series Blu-ray Review (1 Viewer)

[Traveling Matt]

I don't think anyone should necessarily "throw the baby out with the bathwater", BUT I do think there is an important distinction to be made between standard digital cleanup/processing tools (say like DNR etc) which modify/alter what already exists on the raw scan versus an AI tool that is actually trying to CREATE images/detail of images that DOES NOT EXIST on the film scan. The example in question for this set seems to pretty clearly show that the latter was used, at least on one episode and I would think likely more. Regardless of how great something looks, if some of the image you are watching isn't actually on the film as it was filmed in the 1950's but is instead the result of some 2020's era AI technology "guessing" at what the detail/image is supposed to be, I personally consider that a notable problem, no matter how great it might otherwise make things look (setting aside the "horrible example" scene). But then again, I'm not even a fan of retroactively applied HDR for pre-HDR-era movies either, so maybe I'm being too strict on this.

I don't think you're being too strict. I think you've mostly nailed it. Methods that aide in restoring elements is fine. Altering the finished product is not fine, whether it's DNR erasing pencil lines in cel animation or editing live action faces to differ from what was captured by the camera. "Create" is probably the perfect word.

 

[chrislong2]

Thanks, Chris. I appreciate your reply and your view, although I don’t agree on all counts.

AI software is just that… software. It has no will or intent to try to create anything. Its practitioners, who provide the parameters for the tasks to be done, are the ones with the intent.

Hi, I likewise appreciate your reply and I think I understand what you are saying, but.... I am a longtime software developer. The term "AI" gets bantered around rather broadly including by me, but typically I think when most are referring to AI, they are referring to something being created from a conglomeration of other data (in the net/cloud) by an algorithm that makes various decisions based on that data and constraints. To your point, it can also have lines blured with just other programmed tools with various logic.

In the case of the Havana example, we're talking about the creation of an image that the underlying data on the actual film scan does not support - I don't really care if you want to call it "AI", "an AI-powered tool" or "just a tool", but WHATEVER was used made an incorrect image by inferring incorrectly from the data it looked at. There are tools out there that can take a low-res image and based on that low-res data try and "fill in the blanks" so to speak to make it high res. This can be more local-based decision making or cloud-based based on other images/examples etc, the latter of which I would say you'd have to use the term "AI" for. But either way, the problem is this can be very flawed depending on the image because it is still ultimately guesswork by the algorithm on some level. It is pretty clear to me that what is going on in the Havana example is that whatever tool was used on it tried to "fill in the blanks" of the soft/poor resolution of the background actors and in doing that, it did so incorrectly and unnaturally. It also did so only when the foreground character (Ricky) wasn't too close - as soon as he moved within some pre-determined amount of space the algorithm was using, it would revert to not being applied to that background, which means it would jump back and forth or only partially apply depending on the foreground character, which is just odd.

To your point, I am quite sure none of this was intended, yet it is a side-effect of the fundamental problem of them using this type of a tool in the first place. Again, call it AI or not, but why are they even using a tool that is trying to guess what images should look like, rather than just relying on and showing the images that actually ALREADY EXIST on the film? They have used a tool that is attempting to "fill in" data that doesn't exist on the source film and that is a very questionable decision in my book.

 

[RMajidi]

Hi, I likewise appreciate your reply and I think I understand what you are saying, but.... I am a longtime software developer. The term "AI" gets bantered around rather broadly including by me, but typically I think when most are referring to AI, they are referring to something being created from a conglomeration of other data (in the net/cloud) by an algorithm that makes various decisions based on that data and constraints. To your point, it can also have lines blured with just other programmed tools with various logic.

In the case of the Havana example, we're talking about the creation of an image that the underlying data on the actual film scan does not support - I don't really care if you want to call it "AI", "an AI-powered tool" or "just a tool", but WHATEVER was used made an incorrect image by inferring incorrectly from the data it looked at. There are tools out there that can take a low-res image and based on that low-res data try and "fill in the blanks" so to speak to make it high res. This can be more local-based decision making or cloud-based based on other images/examples etc, the latter of which I would say you'd have to use the term "AI" for. But either way, the problem is this can be very flawed depending on the image because it is still ultimately guesswork by the algorithm on some level. It is pretty clear to me that what is going on in the Havana example is that whatever tool was used on it tried to "fill in the blanks" of the soft/poor resolution of the background actors and in doing that, it did so incorrectly and unnaturally. It also did so only when the foreground character (Ricky) wasn't too close - as soon as he moved within some pre-determined amount of space the algorithm was using, it would revert to not being applied to that background, which means it would jump back and forth or only partially apply depending on the foreground character, which is just odd.

To your point, I am quite sure none of this was intended, yet it is a side-effect of the fundamental problem of them using this type of a tool in the first place. Again, call it AI or not, but why are they even using a tool that is trying to guess what images should look like, rather than just relying on and showing the images that actually ALREADY EXIST on the film? They have used a tool that is attempting to "fill in" data that doesn't exist on the source film and that is a very questionable decision in my book.

A tip of the hat to you from a fellow longtime software developer.

Your points are valid and well taken.

I do believe AI has an important role to play in the future of digital restoration - so long as the machine learning/pattern recognition is channelled towards recognising and applying what’s already in the scanned element. I’m not sure what happened in the Havana segment, but whatever it was that went wrong was clearly not by design, and I suspect it caused a bit of embarrassment in some Paramount quarters when it surfaced.

We have heard from industry experts on these boards for years now about the costs involved in restoration projects. I feel that AI can be a game-changer if used wisely, as post-scan digital cleanup and correction costs can potentially be greatly reduced. We, the film-loving community stand to gain from this, and I’d be loathe to see a movement form among classic film fans to oppose AI on principle.

The movie industry from its inception and in every aspect of the process - including writing, production, marketing, restoration - has been and will always remain a magic act of smoke and mirrors - I love that about it. AI just promises smoke and mirrors on a more sophisticated level. I’m for its prudent usage, so long as its parameters are closely controlled to rein in the ‘creativity’ bit you rightly highlighted.

 

[Robert Crawford]

I appreciate the respectful nature of the last two posters in regard to AI without the need to lecture other posters about how we should feel or act about this issue. Thank you for the civility of your posts exchange without a hint of condescension.

 

[tenia]

As chrislong2 wrote, I'm not sure the issue is with the tool in itself, but what the issues it can generate at the moment. Plenty of digital tools are used during a digital restoration/remastering process, and they're fine as long at what they do doesn't stand out in the final result.
AI can be such a fine tool, but as long as one will be able to notice that it's been used, because it's turning a picture into something that shows it's been used, then it's either not refined enough yet, or the process of using it hasn't been fine-tuned enough yet, and the question is why they're used despite what they'll generate because of these lack of technical finalization.

Hiventy used Topaz AI on select shots of their 4K master of Mission Cléopatre, because some VFXs were low-rez, and required to be upscaled to 4K and their integration in the rest of the picture to be made more seamless. It's been done in a very cautious way : AI has been used for these shots only, and has been used in a way that makes it difficult to notice. These shots look less defined and a bit harsher, but that's pretty much it. In such a case, the use is AI is fine, because it doesn't stand out and is instead seamless.

So it can be done right even with the current limitations of these tools, which seem to require quite a high level of human QC scrutiny to ensure it's getting caught when going wild. One issue is to make sure that movies (or TV shows) don't suffer from either the lack of reliability of this tool at the moment, and/or the lack of QC required to catch it when it goes wild. The other is why AI is being used at all considering what it might do and what is required to ensure it doesn't, when appropriate restoration/remastering workflows and tools already exist and are tried and true.

 

[chrislong2]

We have heard from industry experts on these boards for years now about the costs involved in restoration projects. I feel that AI can be a game-changer if used wisely, as post-scan digital cleanup and correction costs can potentially be greatly reduced. We, the film-loving community stand to gain from this, and I’d be loathe to see a movement form among classic film fans to oppose AI on principle.

Tip of the hat right back to you. You also bring some great points. I don't disagree that there can be a place for AI in the manner you are talking about. Where I get really nervous is when it is used and doing things like trying to create people's faces where there's not enough existing data to do so a la the Havana example. I Love Lucy, including that Havana episode, predominantly exists on 35mm film. 35mm film has the resolution to capture what the filmmakers more or less intended. So why then is some tool, called AI or not, being run to try and generate more image detail that doesn't exist on the film? In the original film, the background characters in that scene are blurry background characters the way they were intended while the foreground characters get all the attention. Then some tool gets run that tries to take those background characters and give them much more detail than exists on the film. That is my concern. Maybe it's just that one episode. Maybe it's just that one episode that anybody noticed (so far) or where it becomes super obvious that such a tool was run. I still would wonder why they are running such a tool in the first place as I don't understand the need for it. With that said, I personally don't think that so far what amounts to one relatively brief example in one episode rises anywhere near the level of soiling the whole set, but I do think it's very disappointing and I do wonder if as more people watch more of the later season shows we might start seeing other problematic examples come forth because such a tool was in the workflow. My personal philosophy is that when a tool starts actually creating images (either out of thin air or based on too limited data) that don't exist in the underlying film, that especially when we're talking about a classic staple of a show like I Love Lucy, that that is where a line gets crossed and should not really be acceptable, and I certainly don't want that kind of thing to become the norm, which is why I think some people get very defensive about the "use of AI". I definitely agree and believe you are right though that we shouldn't just lump everything in one bucket.

 

[ScottRE]

I admit to not knowing how A.I "works." Going by the freebie fun apps I've used, I feel like it's "type in what you want and hit the generate button" sort of thing. I have to believe it's more involved than that, but someone with more knowledge can chime in there.

The "I Love Lucy" clip bothers me because they trusted the A.I. program enough to skip the QC of the final result. However, over time the tech will improve to probably make it a more consistent tool. And if it does work well enough and saves time and money for old film restoration, maybe it will make releasing of classic films and TV more cost effective.

But damn, they have some bugs to work out.

I wouldn't cancel my order based on this glitch any more than I returned my Star Trek The Motion Picture Directors Edition 4k because of the really awful briefing room scene.

 

[chrislong2]

I appreciate the respectful nature of the last two posters in regard to AI without the need to lecture other posters about how we should feel or act about this issue. Thank you for the civility of your posts exchange without a hint of condescension.

Sometimes us forum members can surprise. lol

 

[Nelson Au]

I admit to not knowing how A.I "works." Going by the freebie fun apps I've used, I feel like it's "type in what you want and hit the generate button" sort of thing. I have to believe it's more involved than that, but someone with more knowledge can chime in there.

The "I Love Lucy" clip bothers me because they trusted the A.I. program enough to skip the QC of the final result. However, over time the tech will improve to probably make it a more consistent tool. And if it does work well enough and saves time and money for old film restoration, maybe it will make releasing of classic films and TV more cost effective.

But damn, they have some bugs to work out.

I wouldn't cancel my order based on this glitch any more than I returned my Star Trek The Motion Picture Directors Edition 4k because of the really awful briefing room scene.

View attachment 239261

Scott, my comment really should be in the Star Trek thread, but I agree that while Daren Docterman is very pleased with revising that shot with the larger windows, and it does follow the original design of where that scene should take place, the revised imagery in the background isn’t as well realized as it could. I don’t think any AI was used, I think it was done the old fashioned way. How quaint. It was a nice try. I still love that new Directors Edition.

 

[ScottRE]

Scott, my comment really should be in the Star Trek thread, but I agree that while Daren Docterman is very pleased with revising that shot with the larger windows, and it does follow the original design of where that scene should take place, the revised imagery in the background isn’t as well realized as it could. I don’t think any AI was used, I think it was done the old fashioned way. How quaint. It was a nice try. I still love that new Directors Edition.

Yeah I never meant to imply that it was AI in the Directors Edition - just that it was a less than successful "enhancement" which I don't like, but didn't prevent me from buying the set and watching it more than once.

 

[mark27b]

Paramount to Tweak ‘I Love Lucy’ Blu-ray After Viral Restoration Fail

Paramount to Tweak ‘I Love Lucy’ Blu-ray After Viral Restoration Fail

Paramount will fix a scene in the remastered version of the 'I Love Lucy' Blu-Ray after some bizarre "over-enhanced" images went viral.

www.indiewire.com