TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

The AI magic behind Sphere's upcoming 'The Wizard of Oz'

54 点作者 radeeyate大约 1 个月前

11 条评论

invisible大约 1 个月前
They showed a bit of this with a high production video of “how we did it” on Tuesday. The outpainting was amazing, details looked crisp, physics sim was pretty good, and the continuity for actors was good. The amount of tech they had to create and manage for the project is remarkable on its own.<p>There were some perspective and lighting issues, and a few small glitchy artifact moments. They probably showed about 5 minutes of footage from various parts of the movie.
actuallyalys大约 1 个月前
This article is kind of strange to me because it implies celluloid film is low quality, but it seems there’s a rough consensus that you can can scan that size of film in to create good 4K video, if not higher. That’s not enough for the Sphere’s full 16K resolution, but that’s more to do with it having extremely high resolution than the film being “tiny” or grainy.
评论 #43668358 未加载
评论 #43666769 未加载
评论 #43682974 未加载
评论 #43667931 未加载
sema4hacker大约 1 个月前
An actual video example would have helped, but nothing to see here.
评论 #43666762 未加载
评论 #43666504 未加载
AIPedant大约 1 个月前
&gt; Take the moment where the Cowardly Lion first pounces on his soon-to-be companions. The camera pans back and forth between the Scarecrow and Tin Man, with cuts to Dorothy hiding behind a tree in the distance. The experience at Sphere called for keeping all these elements together, in hyper-realistic detail.<p>This sounds <i>awful,</i> as if Google thinks direction and editing are technological limitations rather than artistic choices. This entire project seems fit for a culture that loves content and hates art. If these tasteless jokers ever do this to <i>Stalker</i> I will probably riot. (“Carefully composed shots is for anachronistic dweebs, our society demands an Experience Like No Other.”)<p>&gt; Every change, Hays notes, was made in close collaboration with Warner Bros., to ensure continuity with the spirit of the original.<p>Considering Warner Bros. Discovery’s track record is 3 years of selling out their own directors and writers, I would be more confident if Google were left to their own devices! DeepMind likely has more respect for the film than Warner Bros. management.
评论 #43667299 未加载
评论 #43667260 未加载
WalterBright大约 1 个月前
I try to watch silent films, but I just don&#x27;t like:<p>1. reading dialog cards, especially since I can lip read a bit and that&#x27;s not what the characters are saying<p>2. the truly awful &quot;period&quot; music added<p>3. the grainy black and white<p>I&#x27;ve seen some work on the 1929 film &quot;Wings&quot; that was done in high def, sound effects were added, and some small colorization of scenes.<p>I&#x27;d like to take that further:<p>1. remove graininess and fuzziness with AI, and of course all scratches, etc.<p>2. add foley sound effects<p>3. replace dialog cards with dubbed dialog, lip read if possible<p>4. add a decent music sound track (Like what was done with &quot;Metropolis&quot;)<p>5. colorize it!<p>These are all very doable today, as I&#x27;ve seen each done in a small way.<p>Why should we suffer all the limitations of silent movies? I just want to enjoy it, not suffer.
评论 #43667261 未加载
评论 #43676944 未加载
评论 #43668460 未加载
imhoguy大约 1 个月前
Maybe Jodorowsky could make his Dune adaptation finaly that way? They have enough painted material to feed AI.
sema4hacker大约 1 个月前
I wonder if they fixed any of the 147 continuity errors: <a href="https:&#x2F;&#x2F;youtu.be&#x2F;f6yMvD4h5Aw" rel="nofollow">https:&#x2F;&#x2F;youtu.be&#x2F;f6yMvD4h5Aw</a>
mrandish大约 1 个月前
As someone with a background in both film and video production technologies (as well as being a high-end home theater projection enthusiast) I think this project is doomed to be terrible - and not just on aesthetic grounds but also technically (I&#x27;ll explore why it&#x27;s technically doomed in the next paragraphs). But first, what <i>might</i> be good about this? Well, it might be interesting seeing how well they get AI video to maintain consistency of elements between shots. The article is vague but, extrapolating, I assume they probably composited multiple establishing and master shots along with production stills and set schematics to stitch together extended master environment reference plans. Then 3D artists modeled, textured and lit photo-real versions of these extended environments to match the stitched reference material. Those were then used as fine-tuning to constrain the AI outpainting to that ground truth. It might be interesting to see these techniques used to extend the original WoZ 4:3 aspect to a 2.4:1 widescreen format. However, I think it&#x27;s going to be a disaster because they&#x27;re extending the outpainting to encompass the Sphere&#x27;s <i>extreme</i> 165 degree horizontal field of view. Pushing that far will derail any aesthetic or technical value the project may have had.<p>The biggest technical problem is a 165 degree FOV is not just ill-suited for theatrical storytelling, it&#x27;s actively harmful because it significantly constrains the compositional and creative choices a director and cinematographer can make. Historically, Hollywood has experimented with a variety of wider (or taller) than typical cinematic formats including Omnivision, Circle-Vision 360, MagnaVision, Cinerama, IMAX and many others. Over many decades of experimentation, it became clear that, for cinematic storytelling, formats up to around 2.5:1 were mostly upside assuming the costs and space could be supported. Extra-wide formats like Cinerama and IMAX had creatively useful upsides but came with some significant downsides which could be minimized with careful handling. Ultra-wide formats like Omnivision, Circle-Vision and, now, Sphere, were primarily useful <i>only</i> for theme parks and short &quot;You Are There&quot;-type features such as Disneyland&#x27;s Circle-Vision Grand Canyon tour. It can be helpful to refer to the reference chart on this page showing the different FOVs in a typical theater overlaid with the SMPTE, THX and 20th Century Fox recommended FOVs. <a href="https:&#x2F;&#x2F;acousticfrontiers.com&#x2F;blogs&#x2F;articles&#x2F;home-theater-viewing-angles-distances-and-sightlines" rel="nofollow">https:&#x2F;&#x2F;acousticfrontiers.com&#x2F;blogs&#x2F;articles&#x2F;home-theater-vi...</a><p>Experimentation showed that ultra-wide formats, which are more than double typical cinematic FOVs and originally developed for world fairs, are simply ill-suited to cinematic storytelling. In addition to audience fatigue during longer run-times, significant technical challenges of optical distortion, and high costs - perhaps the worst part was the director losing much of the ability to signal to the audience what&#x27;s important through framing and composition. This signal channel between the director and audience usually goes unnoticed by most viewers but it&#x27;s a profoundly important storytelling tool for directors and cinematographers. To be fair, I do think immersive&#x2F;surround visual formats <i>can</i> be useful in the context of a theme park attraction, amusement ride, VR headset or interactive gaming. They just don&#x27;t work well for cinematic storytelling - like Wizard of Oz. It&#x27;s a good tool being used for the wrong job.<p>Recently, I screened the &quot;Postcards from Earth&quot; movie at the Sphere. This movie was created specifically to launch the Sphere and is their featured demo. And they indeed struggled mightily with the issues I&#x27;ve outlined. Ultimately, they chose to mostly not use more than about a 60 degree slice from the center of their 165 degree canvas, at least for anything significant to the story. All that very expensive, compromise-causing extended FOV was relegated to ambient scenic support except for a few brief &quot;stunts&quot; where some large object would arrive from overhead or an edge. But even those would quickly move from being at the edges (and too close&#x2F;big), to exist in the center 60 degrees like everything else. Also, Sphere content must strictly limit any camera panning, tilting or side dollying to avoid causing motion sickness.<p>The issues I&#x27;ve outlined above are primarily &quot;Production&quot; problems with Ultra-Wide FOVs, however the extreme format also causes significant &quot;Presentation&quot; problems. These presentation problems come from the Sphere going all-in on creating such a large, extremely wide-angle, wrap-around presentation that fills the entire visual field for 17,000 seats. Unfortunately, choosing that &quot;feature&quot; as the top priority requires other important tech aspects of visual presentation like contrast, dynamic range and resolution to be significantly compromised. A key problem is that the wrap-around screen being 165 degrees causes it to illuminate itself nuking the contrast. The sides down near the horizon line are opposite and shining directly at each other. Another significant issue is that it&#x27;s almost impossible to shoot or present real-world camera content able to fill the entire 165 degree Sphere screen with a single natural image. As near as I could tell, the entire Postcards from Earth movie doesn&#x27;t contain even a single full-screen frame that was shot with one camera. It&#x27;s all CGI with occasional real-world camera shots composited into small frames within the wrap-around CGI field. This is because it&#x27;s incredibly challenging (if not entirely impossible) to photograph a single image that wide and tall while keeping the perspective from being severely distorted. During the &quot;Planet Earth&quot;-type scenes, the director clearly had to go to great effort to keep any real-world object with straight lines from getting too big. On top of the significant cinematic, compositional and tech issues, the content of Postcards from the Earth is also weak. The story was trite, shop-worn and heavy-handed. The acting, music, cinematography, etc was overall weak - basically what I&#x27;d call &quot;pretty good for a video game cut scene but certainly not AAA cinema grade.&quot;<p>In terms of nice things to say... well, the audio presentation wasn&#x27;t bad. By which I mean, it wasn&#x27;t great but it was impressively good for that huge of a space dominated by a massive non-parabolic reflector. Basically, the massive size and unusual shape of the space make it extremely challenging to provide a decent audio field to the majority of seats. IMHO, the audio engineering team over-achieved in the degree to which they&#x27;re able to address many of those challenges. There are a huge number of tuned speakers hidden behind the acoustically semi-transparent screen driven by a lot of DSP power. Very expensive and technically quite difficult. Unfortunately, the resulting audio - while technically impressive given the significant constraints, still isn&#x27;t nearly as good as a well-tuned flagship Dolby Atmos, THX-certified theater in Hollywood or Manhattan. Another unfortunate aspect is the perforations in the screen required to enable the audio transparency further reduce the screen&#x27;s ability to generate peak nits of brightness.<p>If you want to experience today&#x27;s highest fidelity theatrical imagery for cinematic storytelling, the Sphere isn&#x27;t it. Sadly, it&#x27;s not even close (which makes the &gt;$100 price for a bad movie, presented poorly especially egregious). The best you can experience today is visiting one of the 31 real IMAX cinemas (out of 1700 IMAX branded screens (aka &quot;LieMAX&quot; screens)) but only when they are showing a movie A) Distributed in the full 15 perf &#x2F; 70mm 2D IMAX format, and which was B) Specifically shot to utilize the full 1.43:1 aspect ratio of the full IMAX format. Unfortunately, that&#x27;s a small minority of what&#x27;s shown on those 31 screens. Most films shown on IMAX screens weren&#x27;t really shot specifically for full 1.43 IMAX format (which is super tall compared to normal cinema aspect), so they&#x27;re just open gate (unmasked) versions of films primarily framed and lit for typical wide format exhibition (like 2.39:1). Also, avoid anything shown in 3D because the 3D projection process (whether IMAX or not) always significantly reduces brightness, contrast and resolution vs 2D projection.
评论 #43668610 未加载
评论 #43668587 未加载
Xcelerate大约 1 个月前
&gt; “When the request came to us, I was almost jumping up and down,” says Dr. Irfan Essa, a principal research scientist at Google DeepMind and director of its Atlanta lab.<p>Wait, what? When did Google DeepMind open an Atlanta office? That seems like a news story in and of itself... maybe I’ve been under a rock the last few years.
评论 #43670266 未加载
leptons大约 1 个月前
AI, huh? So will Dorothy have 5 fingers or 8 fingers on each hand?
评论 #43668044 未加载
评论 #43666608 未加载
ge96大约 1 个月前
The NASCAR layout was pretty nuts and Anema (EDM) robot guy leaps off a cliff