Goading companies into improving image and video generation by showing them how terrible they are is only going to make them go faster, and personally I’d like to enjoy the few moments I have left thinking that maybe something I watch is real.
It will evolve into people hooked into entertainment suits most of the day, where no one has actual relationships or does anything of consequence, like some sad mashup of Wall-E and Ready Player One.
If we’re lucky, some will want to meatspace with augmented reality.
Maybe we’ll get really nice holovisions, where we can chat with virtual celebrities.
Who needs that?
We’re already having to shoot up weight-loss drugs because we binge watch streaming all the time. We’ve all given up, assuming AI will do everything. What good will come from having better technology when technology is already doing harm?
I am looking at ways to approximate Gaussian splats without having to reinvent the wheel, but I'm a bit over my depth since I haven't been playing a lot of attention to those in general.
I'm quite delighted that the gif banding artefacts make it look life the photi of a fire is flickering, and also highly impressed that the AI was able to recognize the fire as a photo within a photo and keep it in 2d.
I just refactored the rendering and resampling approach. Took me a few tries to figure out how to remove the banding masks from the layers, but with more stacked layers and a bit of GPT-foo to figure out the API it sort of works now (updated the GIF)
Keep in mind that this is not Gaussian splat rendering but just a hacked approximation--on my NVIDIA machine that looks way smoother.
Can someone ELI5 what this does? I read the abstract and tried to find differences in the provided examples, but I don't understand (and don't see) what the "photorealistic" part is.
Imagine history documentaries where they take an old photo and free objects from the background and move them round giving the illusion of parallax movement. This software does that in less than a second, creating a 3D model that can be accurately moved (or the camera for that matter) in your video editor. It's not new, but this one is fast and "sharp".
Oh man. I never thought about how Ken Burns might use that.
Already you sometimes see where manually cut out a foreground person from the background and enlarge them a little bit and create a multi-layer 3D effect, but it's super-primitive and I find it gimmicky.
Bringing actual 3D to old photographs as the camera slowly pans or rotates slightly feels like it could be done really tastefully and well.
Takes a 2D image and allows you to simulate moving the angle of the camera with correct-ish parallax effect and proper subject isolation (seems to be able to handle multiple subjects in the same scene as well)
I guess this is what they use for the portrait mode effects.
It turns a single photo into a rough 3D scene so you can slightly move the camera and see new, realistic views. "Photorealistic" means it preserves real textures and lighting instead of a flat depth effect. Similar behavior can be seen with Apple's Spatial Scene feature in the Photos app: https://files.catbox.moe/93w7rw.mov
From a single picture it infers a hidden 3D representation, from which you can produce photorealistic images from slightly different vantage points (novel views).
Basically depth estimation to split the scene into various planes, and then inpainting to work out the areas in the obscured parts of the planes, and then the free movement of them to allow for parallax. Think of 2D side scrolling games that have various different background depths to give illusion of motion and depth.
Apple does something similar right now in their photos app, generating spatial views from 2d photos, where parallax is visible by moving your phone. This paper’s technique seems to produce them faster. They also use this same tech in their Vision Pro headset to generate unique views per eye, likewise on spatialized images from Photos.
Black Mirror episode portraying what this could do: https://youtu.be/XJIq_Dy--VA?t=14. If Apple ran SHARP on this photo and compared it to the show, that would be incredible.
Agreed, this is a terrible presentation. The paper abstract is bordering on word salad, the demo images are meaningless and don’t show any clear difference to the previous SotA, the introduction talks about “nearby” views while the images appear to show zooming in, etc.
I note the lack of human portraits in the example cases.
My experience with all these solutions to date (including whatever apple are currently using) is that when viewed stereoscopically the people end up looking like 2d cutouts against the background.
I haven't seen this particular model in use stereoscopically so I can't comment as to its effectiveness, but the lack of a human face in the example set is likely a bit of a tell.
Granted they do call it "Monocular View Synthesis", but i'm unclear as to what its accuracy or real-world use would be if you cant combine 2 views to form a convincing stereo pair.
Im not sure how the depth estimation alone translates into the view synthesis, but the current implementation on-device is definitely not convincing for literally any portrait photographs I have seen.
True stereoscopic captures are convincing statically, but don't provide the parallax.
Apple's Spatial Scene in the Photos app shows similar behavior, turning a single photo into a small 3D scene that you can view by tilting the phone. Demo here: https://files.catbox.moe/93w7rw.mov
This would be really fun to create stereoscopic videos with. Take a video input, offset x+0.5 or some coefficient, take the output, put them side by side (or interlaced for shutter glasses) and viola! 3D movies.
I could not find any mention of it but does this use regenerative AI? I can’t imagine it able to accomplish anything like this without using a large graphical
Model in the back.
Apple dropping this is interesting. They've been quiet on the flashy AI stuff while everyone else is yelling about transformers, but 3D reconstruction from single images is actually useful hardware integration stuff.
What's weird is we're getting better at faking 3D from 2D than we are at just... capturing actual 3D data. Like we have LiDAR in phones already, but it's easier to neural-net your way around it than deal with the sensor data properly.
Five years from now we'll probably look back at this as the moment spatial computing stopped being about hardware and became mostly inference. Not sure if that's good or bad tbh.
In Chapter D.7 they describe: "The complex reflection in water is interpreted by the network as a distant mountain, therefore the water surface is broken."
This is really interesting to me because the model would have to encode the reflection as both the depth of the reflecting surface (for texture, scattering etc) as well as the "real depth" of the reflected object. The examples in Figure 11 and 12 already look amazing.
This is incredibly cool. It's interesting how it fails in the section where you need to in-paint. SVC seems to do that better than all the rest, though not anywhere close to the photorealism of this model.
Is there a similar flow but to transform either a video/photo/NeRF of a scene into a tighter, minimal polygon approximation of it. The reason I ask is that it would make some things really cool. To make my baby monitor mount I had to knock out the calipers and measure the pins and this and that, but if I could take a couple of photos and iterate in software that would be sick.
You'd still need one real measurement at least: this might get proportions right if background can be clearly separated, but the absolute size of an object can be worlds apart.
That's cool and all, but it seems like only the first step in this, where they go from 2D photo all the way to fully animated (animatable?) characters: https://www.youtube.com/watch?v=DSRrSO7QhXY
Have a look through the rest of the images. TMPI has some pretty obvious shortcomings in a lot of them.
1. Sky looks jank
2. Blurry/warped behind the horse
3. The head seems to move a lot more than the body. You could argue that this one is desirable
4. Bit of warping and ghosting around the edges of the flowers. Particularly noticeable towards the top of the image.
5. Very minor but the flowers move as if they aren't attached to the wall
I'm confused, does it actually generate environments from photographs? I can't view the galleries since I didn't sign up for emails but all of the gallery thumbnails are AI, not photos.
Works great, model file is 2.8 GB, on M2 rendering took a few seconds, result is guassian .ply file but repo implementation requires CUDA card to render video, I have used one of webgl live renderers from here https://github.com/scier/MetalSplatter?tab=readme-ov-file#re...
That is really impressive. However, it was a bit confusing at first because in the koala example at the top, the zoomed in area is only slightly bigger than the source area. I wonder why they didn't make it 2-3x as big in both axes like they did with the others.
I understand AI for reasoning, knowledge, etc. I haven't figured out how anyone wants to spend money for this visual and video stuff. It just seems like a bad idea.
Simulation. It takes a lot of effort today to bring up simulations in various fields. 3 D programming is very nontrivial and asset development is extremely expensive. If I have a workspace I can take a photo of and just use it to generate a 3d scene I can then use it in simulations to test ideas out. This is particularly useful in robotics and industrial automation already.
I don't see any examples of a 3D scene information usable for simulation. If you want to simulate something hitting a table, you need the whole table (surface) in space, not just some spatial illusion effect extrapolated from an image of a table. I also think modelling the 3D objects for simulation is the least expensive part of an simulation... the simulation is the expensive thing.
I doubt this will be useful for robotics or industrial automation, where you need an actual spatial, or functional understanding of the object/environment.
Photo apps on phones (can you still call them cameras?) already have a lot of "AI" to enhance photos and videos taken. Some of it is technological necessity, since you're capturing something through a tiny hole, a lot of it is sexying it up to appeal to people, because hey, people would prefer a cinema-quality depiction of their memories rather than the reality...
This specific paper is pretty different to the kind of photo/video generation that has been hyped up in recent years. In this case, I think this might be what they're using for the iOS spatial wallpaper feature, which is arguably useless but is definitely an aesthetic differentiator to Android devices. So, it's indirectly making money.
Do people not spend on entertainment? Commercials? It's probably less of a bad idea than knowledge. AI giving a bad visual has less negatives than giving the wrong knowledge leading to the wrong decision.
"Unsplash > Gen3C > The fly video" is nightmare fuel. View at your own risk: https://apple.github.io/ml-sharp/video_selections/Unsplash/g...
Goading companies into improving image and video generation by showing them how terrible they are is only going to make them go faster, and personally I’d like to enjoy the few moments I have left thinking that maybe something I watch is real.
It will evolve into people hooked into entertainment suits most of the day, where no one has actual relationships or does anything of consequence, like some sad mashup of Wall-E and Ready Player One.
If we’re lucky, some will want to meatspace with augmented reality.
Maybe we’ll get really nice holovisions, where we can chat with virtual celebrities.
Who needs that?
We’re already having to shoot up weight-loss drugs because we binge watch streaming all the time. We’ve all given up, assuming AI will do everything. What good will come from having better technology when technology is already doing harm?
It turns out the Great Filter is that any species with the technology to colonize space also has the technology to soma itself into annihilation.
https://en.wikipedia.org/wiki/Great_Filter
1 reply →
Early AI „everything turns into dog heads“ vibes. Beautiful.
I miss those. Anyone know if it's still possible to get the models etc. needed to generate them?
5 replies →
san check, 1d10
Seth Brundle has entered the chat.
Well, I got _something_ to work on Apple Silicon:
https://github.com/rcarmo/ml-sharp (has a little demo GIF)
I am looking at ways to approximate Gaussian splats without having to reinvent the wheel, but I'm a bit over my depth since I haven't been playing a lot of attention to those in general.
I'm quite delighted that the gif banding artefacts make it look life the photi of a fire is flickering, and also highly impressed that the AI was able to recognize the fire as a photo within a photo and keep it in 2d.
The example doesn't look particularly impressive to say the least. Look at the bottom 20 %
I just refactored the rendering and resampling approach. Took me a few tries to figure out how to remove the banding masks from the layers, but with more stacked layers and a bit of GPT-foo to figure out the API it sort of works now (updated the GIF)
Keep in mind that this is not Gaussian splat rendering but just a hacked approximation--on my NVIDIA machine that looks way smoother.
1 reply →
Can someone ELI5 what this does? I read the abstract and tried to find differences in the provided examples, but I don't understand (and don't see) what the "photorealistic" part is.
Imagine history documentaries where they take an old photo and free objects from the background and move them round giving the illusion of parallax movement. This software does that in less than a second, creating a 3D model that can be accurately moved (or the camera for that matter) in your video editor. It's not new, but this one is fast and "sharp".
Gaussian splashing is pretty awesome.
Oh man. I never thought about how Ken Burns might use that.
Already you sometimes see where manually cut out a foreground person from the background and enlarge them a little bit and create a multi-layer 3D effect, but it's super-primitive and I find it gimmicky.
Bringing actual 3D to old photographs as the camera slowly pans or rotates slightly feels like it could be done really tastefully and well.
What are free objects?
9 replies →
Takes a 2D image and allows you to simulate moving the angle of the camera with correct-ish parallax effect and proper subject isolation (seems to be able to handle multiple subjects in the same scene as well)
I guess this is what they use for the portrait mode effects.
It turns a single photo into a rough 3D scene so you can slightly move the camera and see new, realistic views. "Photorealistic" means it preserves real textures and lighting instead of a flat depth effect. Similar behavior can be seen with Apple's Spatial Scene feature in the Photos app: https://files.catbox.moe/93w7rw.mov
From a single picture it infers a hidden 3D representation, from which you can produce photorealistic images from slightly different vantage points (novel views).
There's nothing "hidden" about the 3d represenation. It's a point cloud (in meters) with colors, and a guess at the the "camera" that produced it.
(I am oversimplifying).
3 replies →
Basically depth estimation to split the scene into various planes, and then inpainting to work out the areas in the obscured parts of the planes, and then the free movement of them to allow for parallax. Think of 2D side scrolling games that have various different background depths to give illusion of motion and depth.
Apple does something similar right now in their photos app, generating spatial views from 2d photos, where parallax is visible by moving your phone. This paper’s technique seems to produce them faster. They also use this same tech in their Vision Pro headset to generate unique views per eye, likewise on spatialized images from Photos.
It makes your picture 3D. The "photorealistic" part is "it's better than these other ways".
Black Mirror episode portraying what this could do: https://youtu.be/XJIq_Dy--VA?t=14. If Apple ran SHARP on this photo and compared it to the show, that would be incredible.
Or if you prefer Blade Runner: https://youtu.be/qHepKd38pr0?t=107
One more example from Star Trek Into Darkness https://youtu.be/p7Y4nXTANRQ?t=61
Agreed, this is a terrible presentation. The paper abstract is bordering on word salad, the demo images are meaningless and don’t show any clear difference to the previous SotA, the introduction talks about “nearby” views while the images appear to show zooming in, etc.
I note the lack of human portraits in the example cases.
My experience with all these solutions to date (including whatever apple are currently using) is that when viewed stereoscopically the people end up looking like 2d cutouts against the background.
I haven't seen this particular model in use stereoscopically so I can't comment as to its effectiveness, but the lack of a human face in the example set is likely a bit of a tell.
Granted they do call it "Monocular View Synthesis", but i'm unclear as to what its accuracy or real-world use would be if you cant combine 2 views to form a convincing stereo pair.
They're using their Depth Pro model for depth estimation, and that seems to do faces really well.
https://github.com/apple/ml-depth-pro
https://learnopencv.com/depth-pro-monocular-metric-depth/
Im not sure how the depth estimation alone translates into the view synthesis, but the current implementation on-device is definitely not convincing for literally any portrait photographs I have seen.
True stereoscopic captures are convincing statically, but don't provide the parallax.
2 replies →
cuda gpu only
https://github.com/apple/ml-sharp#rendering-trajectories-cud...
The gaussian splat output can be generated with CPU (this was honestly one of the easiest AI repos to get running).
Interestingly Apple’s own models don’t work on MPS. Well, I guess you just have to wait for few years..
Fixed that: https://github.com/rcarmo/ml-sharp
This is specifically only for video rendering. The model itself works across GPU, CPU, and MPS.
No, model works without CUDA then you have .ply that you can drop into gaussian splatter viewer like https://sparkjs.dev/examples/#editor
CUDA is needed to render side scrolling video, but there is many ways to do other things with result.
> photorealistic 3D representation from a single photograph in less than a second
Apple's Spatial Scene in the Photos app shows similar behavior, turning a single photo into a small 3D scene that you can view by tilting the phone. Demo here: https://files.catbox.moe/93w7rw.mov
It‘s awful and often creates a blurry mess in the imaginated space behind the object.
Photoshop content aware fill could do equally or better many years ago.
This would be really fun to create stereoscopic videos with. Take a video input, offset x+0.5 or some coefficient, take the output, put them side by side (or interlaced for shutter glasses) and viola! 3D movies.
Is there a link with some sample gaussian splat files coming from this model? I couldn't find it.
Without that that it's hard to tell how cherry-picked the NVS video samples are.
EDIT: I did it myself, if anyone wants to check out the result (caveat, n=1): https://github.com/avaer/ml-sharp-example
Impressive but something doesn't feel right to me.. Possibly too much sharpness, possibly a mix of cliches, all amplified at once.
For me, TMPI and SHARP look great. TMPI is consistently brighter, though, with me having no clue which is more correct.
I could not find any mention of it but does this use regenerative AI? I can’t imagine it able to accomplish anything like this without using a large graphical Model in the back.
This seems like what they have been doing with album covers on applemusic for a couple years.
Apple dropping this is interesting. They've been quiet on the flashy AI stuff while everyone else is yelling about transformers, but 3D reconstruction from single images is actually useful hardware integration stuff.
What's weird is we're getting better at faking 3D from 2D than we are at just... capturing actual 3D data. Like we have LiDAR in phones already, but it's easier to neural-net your way around it than deal with the sensor data properly.
Five years from now we'll probably look back at this as the moment spatial computing stopped being about hardware and became mostly inference. Not sure if that's good or bad tbh.
Will include this one in my https://hackernewsai.com/ newsletter.
I wonder if humans are any different. We don't have LIDAR in our eyes but we approximate depth "enough" with only our 2D input
In Chapter D.7 they describe: "The complex reflection in water is interpreted by the network as a distant mountain, therefore the water surface is broken."
This is really interesting to me because the model would have to encode the reflection as both the depth of the reflecting surface (for texture, scattering etc) as well as the "real depth" of the reflected object. The examples in Figure 11 and 12 already look amazing.
Long tail problems indeed.
It would be interesting to see how much better this algorithm would be with a stereo pair as input.
Not only do many VR and AR systems acquire stereo, we have historical collections of stereo views in many libraries and museums.
This is incredibly cool. It's interesting how it fails in the section where you need to in-paint. SVC seems to do that better than all the rest, though not anywhere close to the photorealism of this model.
Is there a similar flow but to transform either a video/photo/NeRF of a scene into a tighter, minimal polygon approximation of it. The reason I ask is that it would make some things really cool. To make my baby monitor mount I had to knock out the calipers and measure the pins and this and that, but if I could take a couple of photos and iterate in software that would be sick.
You'd still need one real measurement at least: this might get proportions right if background can be clearly separated, but the absolute size of an object can be worlds apart.
That's cool and all, but it seems like only the first step in this, where they go from 2D photo all the way to fully animated (animatable?) characters: https://www.youtube.com/watch?v=DSRrSO7QhXY
This is great for turning a photo into a dynamic-IPD stereo pair + allows some head movement in VR.
Ah and the dynamic IPD component preserves scale?
The resulting animations feel more like "Live2D" than 3D.
So this is the secret sauce behind Cinematic mode. The fake bokeh insanity has reached its climax!
As well as their "Spatial Scene" mode for lock screen images, which synthesizes a mild parallax effect as you move the phone.
It's available for everyday photos, portraits, everything, not just lock screens.
you can also press the button while viewing a photo in the Photos app to see this.
So Deckard got lucky that the picture enhancement machine allucinated the correct clue? But that was boundto happen 6 years ago, no AI yet.
Enhance! https://www.youtube.com/watch?v=LhF_56SxrGk
I thought this was going to be the Super Troopers version
TMPI looks just as good if not better.
Have a look through the rest of the images. TMPI has some pretty obvious shortcomings in a lot of them.
1. Sky looks jank 2. Blurry/warped behind the horse 3. The head seems to move a lot more than the body. You could argue that this one is desirable 4. Bit of warping and ghosting around the edges of the flowers. Particularly noticeable towards the top of the image. 5. Very minor but the flowers move as if they aren't attached to the wall
Disagree - look at the sky in the seaweed shot. It doesn't quite get the depth right in anything, and the edges of things look off.
Agreed. The head of the fly also seems to have weird depth.
Quite cool!
The paper is just a word salad and it's not better than previous sota? I might be missing a key element here.
I want to see with people
See also Spaitial[0] which announced today full 3D environment generation from a single image
[0]https://www.spaitial.ai/
Requires email to view anything, that’s sad
I'm confused, does it actually generate environments from photographs? I can't view the galleries since I didn't sign up for emails but all of the gallery thumbnails are AI, not photos.
> I'm confused, does it actually generate environments from photographs?
It’s a website that collects people’s email addresses
The best I've seen so far is Marble from World Labs, though that gives you a full 360 environment and takes several minutes to do so.
Why are all their examples of rooms?
Why no landscape or underwater scenes or something in space, etc.?
Constrained environments are much simpler.
I believe this company is doing image (or text) -> off the shelf image model to generate more views -> some variant of gaussian splatting.
So they aren't really "generating" the world as one might imagine.
Works great, model file is 2.8 GB, on M2 rendering took a few seconds, result is guassian .ply file but repo implementation requires CUDA card to render video, I have used one of webgl live renderers from here https://github.com/scier/MetalSplatter?tab=readme-ov-file#re...
That is really impressive. However, it was a bit confusing at first because in the koala example at the top, the zoomed in area is only slightly bigger than the source area. I wonder why they didn't make it 2-3x as big in both axes like they did with the others.
[dead]
[flagged]
I understand AI for reasoning, knowledge, etc. I haven't figured out how anyone wants to spend money for this visual and video stuff. It just seems like a bad idea.
Simulation. It takes a lot of effort today to bring up simulations in various fields. 3 D programming is very nontrivial and asset development is extremely expensive. If I have a workspace I can take a photo of and just use it to generate a 3d scene I can then use it in simulations to test ideas out. This is particularly useful in robotics and industrial automation already.
I don't see any examples of a 3D scene information usable for simulation. If you want to simulate something hitting a table, you need the whole table (surface) in space, not just some spatial illusion effect extrapolated from an image of a table. I also think modelling the 3D objects for simulation is the least expensive part of an simulation... the simulation is the expensive thing.
I doubt this will be useful for robotics or industrial automation, where you need an actual spatial, or functional understanding of the object/environment.
1 reply →
Photo apps on phones (can you still call them cameras?) already have a lot of "AI" to enhance photos and videos taken. Some of it is technological necessity, since you're capturing something through a tiny hole, a lot of it is sexying it up to appeal to people, because hey, people would prefer a cinema-quality depiction of their memories rather than the reality...
This specific paper is pretty different to the kind of photo/video generation that has been hyped up in recent years. In this case, I think this might be what they're using for the iOS spatial wallpaper feature, which is arguably useless but is definitely an aesthetic differentiator to Android devices. So, it's indirectly making money.
Do people not spend on entertainment? Commercials? It's probably less of a bad idea than knowledge. AI giving a bad visual has less negatives than giving the wrong knowledge leading to the wrong decision.