[ad_1]
It’s a coincidence of course, but the day Metal Gear Solid creator Hideo Kojima muses on Twitter that AI will eventually resume remastering old games, Digital Foundry is releasing a video that uses machine learning to refresh the classic MGS2 reveal of E3 2000. For my money, the original is one of the greatest trailers of all time – and for a number of reasons the opportunities to see this video the way it was meant to be seen are limited, so extract the best archive version there is and update the scale of AI was irresistible.
The trailer itself is nothing short of amazing, even today – and for a number of reasons. It was first and foremost a real showcase in real time. It was the PlayStation 2 that produced visuals of a quality we had never seen before – real “next-gen” elements were on display even before the machine was released in Europe. Second, the technology on display wasn’t just about graphics, it was about systems. During the nine minutes of intense action, we got to see environmental destruction systems, AI, stealth, specific damage to certain parts of the body and much more. And finally, there was the steering, with cinematic cameras set up to capture the action: the MGS trailer had undertones of James Cameron and Michael Bay in its presentation and likely a lot more influences from Kojima. Yes, there were hints to the story as well, but on top of that Kojima wasn’t afraid to insert humor too. The trailer blew us away with its content, then left us with a smile on our face at the end.
A look at what the AI upconversion does to the original first E3 2000 Metal Gear Solid 2 trailer – includes analysis and full assets, with captions.
Of course, the fact that the trailer came out in the summer of 2000 meant that distributing a decent-quality version was problematic, certainly online. The internet had no decent video playback platform at the turn of the century – and YouTube is full of horrific low-resolution, low-frame-rate renditions of the original asset to date. I was working on a PlayStation magazine at the time and Konami shared a Beta-SP version of the trailer with us for our cover mounted DVDs, but the tape was flipped and a converted PAL render is all that it takes. we stay.
Fortunately, however, a good quality version of the active exists, albeit in very limited quantities. Konami himself realized that this was something special with this trailer and released it on DVD in Japan, so that we can finally enjoy it at its 60fps. ‘origin. My colleague John Linneman lovingly treated the asset and shared it on YouTube in 2017 – and this DVD is the basis for the enhanced version of AI.
Topaz Video Enhance AI is the tool I used for the job, and the task it faces is daunting. We have a 480i source, which means that to create a 4K60 version, the deep learning model has to work with only 240 lines of source frame per frame. Recent versions of the tool have included processing models designed for interlaced sources – two of them, in fact. One is designed for interlaced material from low quality sources like VHS tapes while the other is for DVDs and 1080i tapes.
DF Retro’s take on the iconic MGS2 is mandatory if you’re the least bit interested in the game.
The results are certainly intriguing: the text is easily enlarged, as are the HUD elements seen when Snake uses his tranquilizer gun. Most spectacular is the model’s treatment of Yoji Shinkawa’s iconic work, shown later in the trailer. Elsewhere, the results are mixed but still interesting. The internal areas shown in the trailer are skillfully managed, and it’s actually quite remarkable how the external shots in the storm see the detail increase dramatically – perhaps too much, to the point of being too fine-tuned at times. . Things like soldier fatigue that gain clarity and definition are easy to deal with, but the way the AI model finds detail in the storm can challenge the imagination. The biggest limitation is that the MGS2 was originally delivered to the field at the time – never delivering more than 240 lines per frame. The AI model isn’t trained to smooth out the most egregious irregularities and in some scenes there is little gain – just extra clarity.
However, as Kojima says, the future is in AI and tools like Topaz Video Enhance AI are already in use in the industry, especially being deployed for game remasters. Textures are ‘remastered’ with AI, while animated video footage is rendered at considerably lower resolution thanks to AI scaling. The cleaner the original source, the more an AI is able to do. But there is some AI to see to matter. The outputs can also be random: in my experiments, what I have found the most limiting is that the actual adjustment of the output image is limited to nonexistent, and the models themselves are like “boxes.” black ”. As a user, we don’t understand how they do what they do and therefore how to improve their results if something goes wrong. In the case of my efforts here, I can’t help but wonder how much better the results would be if the training data came entirely from video games.
Ultimately, the technology is extremely exciting – but we’re only in the early days and there’s a long way to go. Deep learning has a way to go not only in terms of intelligence, but also in terms of speed: the full trailer is around nine minutes long but took around three hours to process. It is with AI study material without additional help. Nvidia DLSS has shown that a more targeted application linked directly to a game engine produces faster results by several orders of magnitude and better – and in the short to medium term, this is where we’ll probably see the most interesting and dramatic examples of how deep learning can drastically improve the quality of the games we play.
[ad_2]
Source link