• il y a 2 mois
Transcription
00:00Ok, ça peut être la plus impressionnante démonstration de technologie que j'ai jamais vue.
00:09C'est aussi l'une des plus difficiles à expliquer en vidéo, mais je vais essayer.
00:12Donc, il y a quelque chose qui s'appelle Google Project Starline.
00:15Vous l'avez peut-être entendu, vous l'avez peut-être pas.
00:16Mais Google a montré quelques exemples de vidéos de ce live, super immersif,
00:21comme une boîte de vidéo en 3D, où vous vous asseyez devant elle.
00:25Et de l'autre côté, quelqu'un d'autre se situe et, grâce au pouvoir de l'Internet,
00:29on dirait qu'une version extrêmement réaliste d'eux est en fait devant vous.
00:34Ce n'est pas juste une vidéo normale, mais ce n'est pas un hologramme,
00:38c'est quelque chose de différent.
00:40De toute façon, on a eu le premier regard sur Google I.O. il y a quelques années,
00:43ils ont montré ce clip dans le keynote, et personne n'a pu l'enregistrer,
00:47personne n'a même pu l'essayer.
00:48Mais maintenant, ils sont de retour avec une version plus compacte, plus simplifiée,
00:55et j'ai pu l'essayer.
00:57Et l'enregistrer.
00:58Juste un rappel, ce premier clip n'est pas vraiment juste,
01:01parce que l'effet n'est pas très bien traduit sur la caméra.
01:04On y reviendra.
01:05Mais j'ai eu l'occasion d'avoir ma première réaction à ça,
01:08ma première première fois de le voir sur la caméra.
01:13All right.
01:14Let's see Project Starline.
01:22Oh, wow.
01:24Hey.
01:25How are you?
01:26Great.
01:27Good to see you. Welcome to Starline.
01:29Thank you.
01:30This is...
01:32Right?
01:33Yeah, it's a lot more...
01:35Yeah, you and I just saw each other right outside,
01:38but we're in two different rooms,
01:39and we want to make it feel like we left that space
01:42and we came back together in a virtual space
01:45that feels like we're completely in the same spot.
01:47It really looks like I can actually...
01:49It looks like you could drop something on this table right here.
01:52Yeah, yeah.
01:53I've got a little apple over here, so let me show you that too.
01:56You know, you and I are in the same room.
01:58Do you want me to pass this over into your space?
02:00It looks like you could do it.
02:04We're meant to feel like this room is just one room
02:07instead of two separate locations.
02:09It could be anywhere in the world,
02:10so this is something we think is the glimpse
02:12of what communication could look like in the next few years.
02:15That's incredible.
02:16Okay.
02:17I've got to learn more about this.
02:19Okay, so what exactly is happening here?
02:22So I'm sitting in front of the new Starline booth,
02:25is what I'd call it.
02:26Now, the first Starline they ever made was much bigger.
02:29They had these, like, full room setups, essentially,
02:31with a bench and computers, the display,
02:34but also a bunch of cameras and depth sensors placed around you.
02:37This new one feels actually much more refined.
02:40It looks much simpler,
02:42and it kind of feels like a reasonable product.
02:44It's a 65-inch display on a stand
02:47with a smaller barrier over the bottom bezel,
02:50and then there are actually some lights on the back
02:53that point at the wall behind the display
02:55that serve as a key light for me, the person on the call.
02:58And then there are no more depth sensors.
03:00It's just a tidy array of color cameras
03:03and microphones and speakers.
03:05So that's at the top, the left, and the right,
03:07and they use AI to create a depth map
03:09of me and the room I'm in.
03:11And then the magic of this is twofold.
03:14It's the display and the computing happening.
03:18So the display, the part giving you
03:20the actually impressive immersive 3D depth effect,
03:23but then also all the computing happening
03:25to turn me, the person on the video call
03:28and the person on the other side,
03:30into this realistic 3D model
03:33just with the camera information
03:35that you can actually look around with head tracking.
03:39Now, clearly, this doesn't translate well on camera,
03:42which is why Google has understandably been super protective
03:45over anyone trying it even
03:48or getting any photos or videos of it
03:50because it just won't look right.
03:52But I was able to convince Google
03:55to let me try something.
03:57You see, basically, once you first sit down
03:59in the Project Starline booth,
04:01this system has to identify where your face is,
04:03and then you can move around and look at stuff
04:05with head and body tracking.
04:07And of course, that's not gonna work
04:09if you just put a camera in there.
04:11But if you do show it a face and a camera at the same time,
04:14like, say, maybe a cardboard face
04:17where you can stick a camera lens
04:19somehow through that cardboard face,
04:22then it would track and you could...
04:25Eh? See what I'm talking about?
04:27Check this out.
04:29So, what you're seeing right now
04:31is as if you were on a video call with me
04:34in the newest version of Google Starline.
04:36This is the first time anyone outside of Google
04:38has seen this sort of visualization,
04:40which is super cool.
04:41So, it looks pretty reasonable right now, right?
04:43I can hold up an object.
04:45You can see the colors and the shapes
04:47and the lighting and textures on it.
04:49And this is all information compiled
04:51from the array of cameras around me.
04:53And the depth information comes from these regular cameras.
04:56And that's all quite cool.
04:58But as you start moving around,
05:00that's when it starts getting a little more interesting
05:03because I can hold things out,
05:05and you start to create this parallax effect
05:07with the head tracking,
05:09where you can start to literally look around
05:11and inspect the object and look underneath things
05:13as if you're literally in the room with me.
05:16It's super convincing to the actual eye
05:18when you see it on this screen.
05:20Even the background behind me is not real.
05:23It's being composited in, but I'm casting a shadow on it,
05:26and that's being rendered in real time too.
05:28It's a lot.
05:29We can even flip a switch
05:31to show a sort of a topographical map
05:33of specifically the depth information
05:36coming from these cameras.
05:38So, again, no depth cameras specifically being used here,
05:41but the array of color cameras
05:43has all this depth information being mapped,
05:46and that's what all the information from the color cameras
05:49is being sort of projected onto.
05:51There's a lot of processing happening here.
05:53So it all comes together
05:55to form this really impressive real-time...
05:59I mean, I use the word immersion very lightly usually,
06:02but I want to stress that it's very immersive
06:05when you actually get to use it.
06:07So hopefully this helps you visualize.
06:09So that is pretty cool.
06:11But even that doesn't translate perfectly
06:14because the feeling of depth
06:16actually comes from the distance between your eyes.
06:19So a camera lens is one eye,
06:21but if you've ever tried to catch something with one eye closed,
06:24you know that humans perceive depth
06:26by the difference between what you see with your left eye
06:28and what you see with your right eye.
06:30Your brain stitches them together and figures out depth that way.
06:33So even that demo doesn't quite get the full realisticness
06:37of what I saw on camera, but it's pretty close.
06:39The light field display in Project Starline
06:41is literally showing a different image
06:43to my left eye and to my right eye.
06:45So taking advantage of that biological fact
06:48and then letting me actually compute depth on the fly
06:51while doing all the head tracking in real time.
06:53It's actually kind of crazy.
06:55Like if you've ever seen 3D glasses,
06:56you know how when you go see a 3D movie,
06:58like one of the eyes is blue, one of the eyes is red?
07:00That's literally because your left eye will see something mapped for that
07:03and your right eye will see something mapped for that,
07:05and that's how it creates the depth effect.
07:07But this is way better than a 3D movie.
07:09It's so much smoother and the head tracking
07:11and everything that goes with it, and it's so good.
07:14But then the craziest part, the most impressive part
07:17is the computing that's happening,
07:19which is rendering the people using Starline
07:21in real time into these 3D models.
07:24So this is not like the metaverse thing
07:27where the VR headset tracks face movement
07:29and then you can have an eye contact,
07:31face-to-face conversation with someone's cartoon avatar.
07:35But no, no, no, this is taking the actual imaging,
07:38the lighting, the way you actually look,
07:40what you're actually wearing,
07:42and rendering that out in real time in 3D space
07:46and making it look like I'm talking to them through a window.
07:50It's kind of amazing.
07:51The eye contact is so one-to-one, real-looking,
07:56and it really just comes from the fact that
07:57I'm sitting here looking forward
07:59and the person on the other side is also looking forward,
08:02and so we're actually looking into each other's eyes.
08:04So anyway, yeah, it is.
08:07The point is, it's super realistic.
08:09I promise you, when I first did my demo,
08:11the first time I ever saw it,
08:13even knowing that it's 3D,
08:15I still felt like I could reach out
08:17and high-five him or fist-bump him.
08:19He held that apple out, actually,
08:21and to my eyes, it looked like he could just drop the apple
08:24he was holding right on the table in front of me,
08:26which of course meant that I looked stupid from this angle
08:28because I'm just reaching out at nothing,
08:30but we got a bunch of reactions
08:32from some other studio team members
08:34who got to join me for this demo.
08:36I'll link a video that has some of their reactions below.
08:38You should definitely check that out.
08:39So it was really only the slight glitching,
08:41a little bit of edges and fading and stuff like that
08:43that kept it out of uncanny valley territory.
08:46Obviously, the eye contact is one thing
08:48and the 3D effect, even it had spatial audio,
08:51so if I leaned over here,
08:52it would sound like more audio came into one of your ears
08:55and that was responsive,
08:57but things like in between the fingers
09:00or the edges of certain fabrics or my hair especially,
09:03some of that stuff could kind of break it a little bit
09:06and you could tell,
09:07but honestly, I wasn't thinking about that at all.
09:10So you might be thinking kind of the same thing I was
09:12after I finished this demo, which is,
09:14okay, this is super cool, but what is this for?
09:17What is this sort of a tech demo actually useful for?
09:22And the answer is, we don't really know yet.
09:26So at this point in time,
09:27Google has worked with a few companies, Salesforce,
09:30WeWork and T-Mobile are some examples
09:32who are literally using some of these booths for meetings,
09:36basically, I guess.
09:37Theoretically, that's better than a Zoom call,
09:40although it still has its limitations,
09:41like the face tracking can only work
09:43with one person at a time,
09:44which means only one person can be in the booth at a time
09:48to get their realistic depth effect,
09:50so adding another person doesn't work.
09:51I think the question will be closer to being answered
09:54when the tech gets even better,
09:56like it's already gone from the size of a room
09:59to the size of like an easel,
10:01it could just be the size of a TV with a backpack on it,
10:04but this will get more realistic,
10:06cheaper, simpler, and just better.
10:09But you know, here's the thing,
10:10I actually don't know that regular people
10:13will actually want this, like hear me out.
10:16If I know anything about bleeding edge tech,
10:18which I've tried a lot of,
10:19it's that regular people, the masses,
10:22are not very fast to pay extra for higher fidelity,
10:26like better quality.
10:27So the cutoff for acceptable quality
10:29is surprisingly low for the masses.
10:31Like think about it, audio quality, I don't know,
10:34streaming on Spotify over Bluetooth
10:36seems to be good enough for most people.
10:37AirPods are the most popular headphones in the world.
10:40You think about cameras and how smartphone cameras
10:43are basically good enough for 99%
10:45of the taking pictures of your kids, you know?
10:48So convenience is king,
10:49that's why right now FaceTime and Zoom
10:52are just a square of a flat low-res video feed on a screen.
10:57That's fine, it's fine for most people.
11:00When you turn it all the way up to Project Starline,
11:03which is like this incredibly realistic,
11:05like you can see micro expressions and textures
11:09and feel like you're in the room with the person,
11:11that's over on one side here,
11:13and FaceTime and Zoom are on the other side here,
11:16and this is where the masses live,
11:18and this is like businesses, you know,
11:21getting a booth so they don't have to fly people out
11:23for lots of meetings overseas,
11:25but it's still too expensive
11:26and too difficult to get to for most people.
11:29But that's just for now.
11:31That's just for this version now.
11:32I'm really looking forward to keeping an eye
11:34on how this tech evolves from Google
11:36and even from others who are working
11:37on this whole 3D light field display technology stuff.
11:41We've actually seen some interesting stuff before.
11:43I actually tried an Asus laptop not too many weeks ago
11:46that does the same thing.
11:48It renders a unique image for each of your eyes
11:51and does head tracking
11:52for an incredibly realistic 3D effect.
11:54But, of course, for that,
11:55you had to look at a certain file
11:56that was built in a certain model
11:58and work with certain software.
11:59This was like rendering real people
12:01and just having a conversation.
12:02It's crazy. It's crazy.
12:03I'm gonna thank you for Google for letting us see this,
12:06and I'm gonna hopefully be able to try future versions
12:09when they get finished,
12:11because this is wild.
12:13Let me know what you guys think.
12:14Thanks for watching.
12:15Catch you in the next one.
12:17Peace.