ai should be able to get forms more solid forms from splats, we need it to recognize which splats need to be a single object through colors. perhaps there is a balance between speed and photorealism in combining the techniques. inside the engine we can place invisible objects to we cant walk through objects in game.
I don't fully agree with the GaSp hype - at least not in this point of time. Photogrammetry, for now, will probably remain the main source of photoscanned content used in games & rendering - mostly due to accuracy and how easy it's to optimize / retopo the models. Plus you don't always want to have all the reflection data you get from NeRF or GaSp. Don't get me wrong - GaSp already have some nice use cases (like virtual tours, etc.), but it's hard for me to see it as a standard for gamedev due to it's volumetric nature. Maybe in the future when we will be able to more accurately turn GaSp volumes into meshes, it will become a more viable option, but for now it's more of a fun gimmick that people love on LinkedIn, rather than a solution to any real existing problem in CGI. It's like: - Whoa, it looks so coo! What can you do with it? - Well... nothing really, at least not yet. - Oookaaay... But! I'm only human and can be mistaken - and maybe I just don't see the whole picture.
@@hasoevo ok, but I'm incorrect how? I've got Luma, I've got Polycam - I'm using both for creating marketing visual content in their limited capabilities. I'm also using Reality Capture and 3df Zephyr and lots of other software. And the GaSp solutions are not replacing my CG pipeline yet. Bad Decisions Studios does what I said above - good content that 'sells' on socials, but it's nothing that's production relevant so far (with emphasis on "so far"). I've seen the same pattern with NeRF - everyone was so stoked on it (and surely - it is cool tech) and I pointed out some problems that wouldn't allow it to outright get into the main industries... and I got pretty much the same arguments like "you should try it yourself" (I did, as opposed to many people who told me to) or "you should see how Corridor Crew used it", etc. Now it's year later - where are all the NeRF based solutions that were promised? Nowhere, because of the current tech limitations. The same goes for GaSp - it's a cool gimmick to show, with little practical use. I do believe it will change with time, it's too cool of a tech to just disappear. I'm just taking a step back on hype that's around it - that's all :)
I enjoy seeing that we follow many of the same technology channels! Great job of summarizing and correlating their relationships! 👏👏👏
This is what convinced me to finally get KIRI Engine. Pretty cool so far. 😎
do you actually prefer KIRI over Luma ?
Can’t wait to be able to turn 3D rendered scenes directly into Gaussian splats!! (Rather than having to render out many frames and then convert them)
Under the hood it would still render to frames. It needs the rendered pixel data - core of the method.
Is there a way of transfering this to Revit or other BIM sofware for architecture?
Does KIRI accept equirectangular and fisheye images for GS rendering?
ai should be able to get forms more solid forms from splats, we need it to recognize which splats need to be a single object through colors. perhaps there is a balance between speed and photorealism in combining the techniques. inside the engine we can place invisible objects to we cant walk through objects in game.
very nive video as usual. I have a question: could you combine GS with AI tech?
Nice
I don't fully agree with the GaSp hype - at least not in this point of time. Photogrammetry, for now, will probably remain the main source of photoscanned content used in games & rendering - mostly due to accuracy and how easy it's to optimize / retopo the models. Plus you don't always want to have all the reflection data you get from NeRF or GaSp. Don't get me wrong - GaSp already have some nice use cases (like virtual tours, etc.), but it's hard for me to see it as a standard for gamedev due to it's volumetric nature.
Maybe in the future when we will be able to more accurately turn GaSp volumes into meshes, it will become a more viable option, but for now it's more of a fun gimmick that people love on LinkedIn, rather than a solution to any real existing problem in CGI.
It's like:
- Whoa, it looks so coo! What can you do with it?
- Well... nothing really, at least not yet.
- Oookaaay...
But! I'm only human and can be mistaken - and maybe I just don't see the whole picture.
No disrespect, but you are incorrect.
To help get you up to date:
- Luma AI
- Bad Decisions Studio TH-cam channel
@@hasoevo ok, but I'm incorrect how? I've got Luma, I've got Polycam - I'm using both for creating marketing visual content in their limited capabilities. I'm also using Reality Capture and 3df Zephyr and lots of other software. And the GaSp solutions are not replacing my CG pipeline yet. Bad Decisions Studios does what I said above - good content that 'sells' on socials, but it's nothing that's production relevant so far (with emphasis on "so far").
I've seen the same pattern with NeRF - everyone was so stoked on it (and surely - it is cool tech) and I pointed out some problems that wouldn't allow it to outright get into the main industries... and I got pretty much the same arguments like "you should try it yourself" (I did, as opposed to many people who told me to) or "you should see how Corridor Crew used it", etc. Now it's year later - where are all the NeRF based solutions that were promised? Nowhere, because of the current tech limitations. The same goes for GaSp - it's a cool gimmick to show, with little practical use. I do believe it will change with time, it's too cool of a tech to just disappear. I'm just taking a step back on hype that's around it - that's all :)
@@hasoevo he's not wrong its really isnt high enough quality to be used in commercial products
sponsored video
Polycam🤘🔥
Bruh
Luma AI is the leader and the best
copy of a copy of a copy
so nothing new from the last year just an app? lol nice ad, no wonder it was at the end of a 7 min video recapping the last year.
This changes nothing if you need a polygonal object in the scene