Not sure if this belongs here but making of Killzone 3 YT vid:
http://www.youtube.com/watch?v=QfvBIHFex9Y&hd=1
http://www.youtube.com/watch?v=QfvBIHFex9Y&hd=1
Game development presentations - a useful reference
- Thread starter liolio
- Start date
The Diablo 3 art panel is highly recommended to everyone who would like to get an insight to an incredibly talented and professional team's work. And also to people who had issues with the art style or certain elements, because they give very reasonable explanations to why and how they got to this result; and they're also not afraid to show some visuals from prototype versions.
This is not strictly related to game development, but I believe it is still interesting for various reasons - here's a behind the scenes look at our work on the Mass Effect 3 cinematic trailer.
http://www.fxguide.com/featured/cinematics-case-study-mass-effect-3/
Here are some of my thoughts about this stuff and how it relates to realtime graphics:
- There's a significant amount of simulation behind most of the effects work and they're all processing and storage intensive. Note the disk cache sizes for the volumetric fluid (fire/smoke) stuff in particular. This kind of work on Wrath of the Titans required more than a hundred terrabytes (!) of disk space, so we're only scratching the surface here. While there's R&D into utilizing GPUs for the calculations, it's still not good enough most of the time... and there's still the caching part to cut down the iteration times when refining the look.
Games are still using 2D sprites and the jump to 3D volumetrics, even at a lower quality level, looks pretty damn expensive.
There's also cloth and hair, not discussed here because it's considered to be more or less a standard feature - games are making progress here but they still don't get the chance to cheat and fix problems like intersections, frame by frame, only for the camera.
- Matte paintings of all kinds are a very cost effective cheat to create grand vistas even with real camera movement. Simple 2D images are just the start as you can see, it's also possible to project back onto 3D geometry and cards to create a sense of depth that even works in stereo. Interactive cameras limit games in utilizing this technique, significantly increasing production costs. Don't be surprised to see newer games taking away camera control more and more often...
- The HDR lighting, reference image shooting, tone mapping stuff is however something that practically any game can make use of, particularly on the next generation of hardware where I expect frame buffer depths, gamma curves and other hw features to be far better optimized for real-world requirements.
- It's also worth to mention that there's enough convergence on the content creation side to allow us to make use of game development related materials of all kinds in our work, and I'm sure the movie VFX guys should also keep their eyes open.
We've already settled on the Right Way to do things in various aspects of CG, and as games adapt these, the limitations of real time execution and hw features force engineers to be more creative and effective in their implementations, thinking of things an offline CG programmer would never bother to look into.
So, I hope the article was interesting
http://www.fxguide.com/featured/cinematics-case-study-mass-effect-3/
Here are some of my thoughts about this stuff and how it relates to realtime graphics:
- There's a significant amount of simulation behind most of the effects work and they're all processing and storage intensive. Note the disk cache sizes for the volumetric fluid (fire/smoke) stuff in particular. This kind of work on Wrath of the Titans required more than a hundred terrabytes (!) of disk space, so we're only scratching the surface here. While there's R&D into utilizing GPUs for the calculations, it's still not good enough most of the time... and there's still the caching part to cut down the iteration times when refining the look.
Games are still using 2D sprites and the jump to 3D volumetrics, even at a lower quality level, looks pretty damn expensive.
There's also cloth and hair, not discussed here because it's considered to be more or less a standard feature - games are making progress here but they still don't get the chance to cheat and fix problems like intersections, frame by frame, only for the camera.
- Matte paintings of all kinds are a very cost effective cheat to create grand vistas even with real camera movement. Simple 2D images are just the start as you can see, it's also possible to project back onto 3D geometry and cards to create a sense of depth that even works in stereo. Interactive cameras limit games in utilizing this technique, significantly increasing production costs. Don't be surprised to see newer games taking away camera control more and more often...
- The HDR lighting, reference image shooting, tone mapping stuff is however something that practically any game can make use of, particularly on the next generation of hardware where I expect frame buffer depths, gamma curves and other hw features to be far better optimized for real-world requirements.
- It's also worth to mention that there's enough convergence on the content creation side to allow us to make use of game development related materials of all kinds in our work, and I'm sure the movie VFX guys should also keep their eyes open.
We've already settled on the Right Way to do things in various aspects of CG, and as games adapt these, the limitations of real time execution and hw features force engineers to be more creative and effective in their implementations, thinking of things an offline CG programmer would never bother to look into.
So, I hope the article was interesting
Clustered Deferred and Forward Shading
Adds depth partitioning to tiled shading.
Multiple Scattering using Light Propagation Volumes
Adds depth partitioning to tiled shading.
Multiple Scattering using Light Propagation Volumes
Last edited by a moderator:
This was tweeted by Repi today, it's a PDF on RSAA. Hoping to read it on my way home from work. :smile:
http://visual-computing.intel-research.net/publications/papers/2012/rsaa/rsaa.pdf
http://visual-computing.intel-research.net/publications/papers/2012/rsaa/rsaa.pdf
Rendering Tech of Space Marine
http://www.slideshare.net/blindrenderer/rendering-tech-of-space-marinekgc-2011
http://www.slideshare.net/blindrenderer/rendering-tech-of-space-marinekgc-2011
Activision's skin rendering presentation from GDC, 300! pages with lots of images, detailed explanations and such:
http://www.iryoku.com/stare-into-the-future
Highly recommended.
http://www.iryoku.com/stare-into-the-future
Highly recommended.
http://advances.realtimerendering.com/s2013/
Some neat ones in there...
Destiny (Bungie) Rendering Engine: The pdf has the notes, so don't worry too much. Interesting read.
Character Rendering Pipeline of Red Engine 3 (CD Projekt)
:>
Some neat ones in there...
Destiny (Bungie) Rendering Engine: The pdf has the notes, so don't worry too much.
Interesting read.Character Rendering Pipeline of Red Engine 3 (CD Projekt)
:>
Nice plog post with plenty of links on Physically Based Rendering.
http://interplayoflight.wordpress.com/2013/12/30/readings-on-physically-based-rendering/
http://interplayoflight.wordpress.com/2013/12/30/readings-on-physically-based-rendering/
In-depth blog post aboot BioInf lighting setup:
http://solid-angle.blogspot.co.uk/2014/03/bioshock-infinite-lighting.html
---
AC4 @GDC
http://bartwronski.com/2014/03/09/my-upcoming-gdc-2014-presentation/
http://solid-angle.blogspot.co.uk/2014/03/bioshock-infinite-lighting.html
---
AC4 @GDC
http://bartwronski.com/2014/03/09/my-upcoming-gdc-2014-presentation/
Essential Math for Games Programmers
http://www.essentialmath.com/tutorial.htm
http://www.essentialmath.com/tutorial.htm
Interesting presentation on Naughty Dog process and technology by Jason Gregory http://www.gameenginebook.com/SINFO.pdf
It includes some details on their job system on PS3 and PS4 and some hardware information. Accessing the cache from the other Jaguar module is almost as costly as accessing the main RAM ! Confirmation of the 6 core available for games on the PS4 (slide 125).
It includes some details on their job system on PS3 and PS4 and some hardware information. Accessing the cache from the other Jaguar module is almost as costly as accessing the main RAM ! Confirmation of the 6 core available for games on the PS4 (slide 125).
Nice presentation from Crytek on deferred MSAA, DoF etc.
Would love to see their new AA method, combination of SMAA with temporal component similar to TXAA.
http://www.crytek.com/download/Sousa_Graphics_Gems_CryENGINE3.pdf
Would love to see their new AA method, combination of SMAA with temporal component similar to TXAA.
http://www.crytek.com/download/Sousa_Graphics_Gems_CryENGINE3.pdf
Everything about SMAA (and its combinations) is here:
http://www.iryoku.com/smaa/
- PDF article (15 pages, very technical with math stuff) about SMAA with its combinations (combined with temporal AA and/or MSAA) + gpu cost comparison with FXAA/MLAA/MSAA/SSAA notably
- Compiled demo to apply youself SMAA on any image (3D scene too I think), I tried it (2D image), it works great.
- Video comparing mosts AAs vs SMAA in real-time in 2D and 3D, great to see the temporal AA + SMAA in real scenes. And the background music is great!
- Downloads of source code freely available for everyone.
Here one excerpt of the impressively complete article:
MystWalker
Newcomer
Couple of papers on an interesting dynamic GI technique called Modular/Delta Radiance Transfer. (DRT being an extension of MRT)
From a high level, it reminds me of Enlighten. e.g. all static objects both receive and bounce light, but dynamic objects only receive light via an Irradiance volume, and how the indirect light is calculated on a simplified environment.
MRT: http://www.cs.utah.edu/~loos/publications/mrt/
DRT: http://www.cs.utah.edu/~loos/publications/drt/
Video of someone else's implementation: http://www.youtube.com/watch?v=WQIOs6M8gyM
A bunch of papers on various methods for interactive GI. I thought the Multi-view Ambient Occlusion was a clever modification/improvement on SSAO.
http://graphics.cs.aueb.gr/graphics/research_illumination.html
Paper/video on Deep Screen Space. Seeing how most nextgen games are approximating so many lighting phenomena in screenspace, this definitely has a lot of applications. Don't know whether performance would be good enough on nextgen system though.
http://www.mpi-inf.mpg.de/~onalbach/DeepScreenSpace/
No formal paper on this one, but I thought it was an interesting realtime GI solution. It's a "many-lights" technique, but, whereas many of these use a reflective shadowmap, this one uses a lowres cubemap probe, ray-casts each face of the cubemap to map the topography of the room/environment, and then places VPLs (Virtual Point Lights) to simulate the indirect bounces. It's pretty fast and it looks good, because it can do several bounces over several frames. He used to have a Unity Web Player demo in the OP, but it looks like he removed it
.
http://forum.unity3d.com/threads/233919-BG-Light-Engine-(With-Real-Time-GI-amp-Area-Lights)
From a high level, it reminds me of Enlighten. e.g. all static objects both receive and bounce light, but dynamic objects only receive light via an Irradiance volume, and how the indirect light is calculated on a simplified environment.
MRT: http://www.cs.utah.edu/~loos/publications/mrt/
DRT: http://www.cs.utah.edu/~loos/publications/drt/
Video of someone else's implementation: http://www.youtube.com/watch?v=WQIOs6M8gyM
A bunch of papers on various methods for interactive GI. I thought the Multi-view Ambient Occlusion was a clever modification/improvement on SSAO.
http://graphics.cs.aueb.gr/graphics/research_illumination.html
Paper/video on Deep Screen Space. Seeing how most nextgen games are approximating so many lighting phenomena in screenspace, this definitely has a lot of applications. Don't know whether performance would be good enough on nextgen system though.
http://www.mpi-inf.mpg.de/~onalbach/DeepScreenSpace/
No formal paper on this one, but I thought it was an interesting realtime GI solution. It's a "many-lights" technique, but, whereas many of these use a reflective shadowmap, this one uses a lowres cubemap probe, ray-casts each face of the cubemap to map the topography of the room/environment, and then places VPLs (Virtual Point Lights) to simulate the indirect bounces. It's pretty fast and it looks good, because it can do several bounces over several frames. He used to have a Unity Web Player demo in the OP, but it looks like he removed it
.http://forum.unity3d.com/threads/233919-BG-Light-Engine-(With-Real-Time-GI-amp-Area-Lights)
Yes, I've been doing a lot of reading on realtime GI solutions, with nextgen here and all
GDC 2014 slides compilation.
http://www.iguanademos.com/Jare/wp/?p=2763
http://www.iguanademos.com/Jare/wp/?p=2763
Last edited by a moderator:
Crytek has put up their slides on Ryse tech:
http://www.crytek.com/download/2014_03_25_CRYENGINE_GDC_Schultz.pdf
http://www.crytek.com/download/2014_03_25_CRYENGINE_GDC_Schultz.pdf
Similar threads
