question about camera projection/3d models
3 posters
Page 1 of 1
question about camera projection/3d models
hey guys and happy new year
Ive just been watching the awesome vfx breakdowns for Stalingrad by main road post and would like to try some of these techniques for my next piece, however I am a little confused by the scene at 1:25. It shows the wireframe/AO/final scene stages. Now do these guys model the scene and then project a matte painting onto them ? or do they unwrap the buildings and texture them ?
https://www.youtube.com/watch?v=BGA73iBDLMA this is the link
Im planning on doing something with a camera movement like this and would love to know the fastest/best way of doing it before committing to painting/modelling.
Cheers again guys
Ive just been watching the awesome vfx breakdowns for Stalingrad by main road post and would like to try some of these techniques for my next piece, however I am a little confused by the scene at 1:25. It shows the wireframe/AO/final scene stages. Now do these guys model the scene and then project a matte painting onto them ? or do they unwrap the buildings and texture them ?
https://www.youtube.com/watch?v=BGA73iBDLMA this is the link
Im planning on doing something with a camera movement like this and would love to know the fastest/best way of doing it before committing to painting/modelling.
Cheers again guys
Re: question about camera projection/3d models
I'm guessing a bit on how they worked to acheive this, but if i was doing something like this - assuming this is a faithful recreation of an existing location - my approach would something like follows:
- Previs the shot, maybe use google earth geo of the location to figure out roughly what area you will see and will need to model out.
- Extensively photograph the location with multiple bracketed exposures. I would probably do a few 360 spherical shoots for sitiching, and then multiple bracketed exposures of various props to fill out your texture library. Process your photos (i.e. stitch the 360 photos, neutral grade etc).
- Model out the scene using the photos as a guide. Get it as close as humanly possible to the real location.
- Use tone mapping to try and remove as much lighting information as possible from the photos, and begin using what you have to texture up the scene. If you use mari you can line up your photos and project them on as a start and then fill out aything that is missing once you have done that using your remaining photo ref.
- Develop all the lookdev and shaders needed to make it all look photoreal (lots of work in that small sentence!!)
- Set up the lighting and render the shot (lots of work and time again in another small sentence!!)
- Anything that doesn't look right or isn't holding up well, consider fixing as dmp that you project back over the final CG renders.
Of course, it's impossible to advise on what route to take without knowing exactly what you will do. Just a simple thing like having more glass windows / reflective surfaces in or a few more inches of movement / parallax in a shot can be the difference between taking a full dmp approach or a full cg approach. But for their work with lots of destruction sims and interactive lighting from explosions i would imagine that a lot of that is full cg....
Does that help?
N
- Previs the shot, maybe use google earth geo of the location to figure out roughly what area you will see and will need to model out.
- Extensively photograph the location with multiple bracketed exposures. I would probably do a few 360 spherical shoots for sitiching, and then multiple bracketed exposures of various props to fill out your texture library. Process your photos (i.e. stitch the 360 photos, neutral grade etc).
- Model out the scene using the photos as a guide. Get it as close as humanly possible to the real location.
- Use tone mapping to try and remove as much lighting information as possible from the photos, and begin using what you have to texture up the scene. If you use mari you can line up your photos and project them on as a start and then fill out aything that is missing once you have done that using your remaining photo ref.
- Develop all the lookdev and shaders needed to make it all look photoreal (lots of work in that small sentence!!)
- Set up the lighting and render the shot (lots of work and time again in another small sentence!!)
- Anything that doesn't look right or isn't holding up well, consider fixing as dmp that you project back over the final CG renders.
Of course, it's impossible to advise on what route to take without knowing exactly what you will do. Just a simple thing like having more glass windows / reflective surfaces in or a few more inches of movement / parallax in a shot can be the difference between taking a full dmp approach or a full cg approach. But for their work with lots of destruction sims and interactive lighting from explosions i would imagine that a lot of that is full cg....
Does that help?
N
Nick- Posts : 194
Join date : 2013-05-18
Location : London
Page 1 of 1
Permissions in this forum:
You cannot reply to topics in this forum
|
|