Hey Pat,

glad that you liked "HELP", I worked on it as a 2D TD and Compositor. In our 
case we ended up comping a lot in Latlong space. It works for a lot of cases 
and you get used to it pretty quickly. We also had tools that allowed us to 
quickly check things in a rectilinear view(similiar to a cube face) or move it 
from one image space to the other. Additionally we used quite a bit of 
projections which worked really well. Yes shifting it around with a spherical 
transform is also an option we used. Just keep in mind, that any transformation 
will most likely be a filter hit, and therefore soften your plate.

cheers,
Patrick

On 28.06.2015, at 07:21, Pat Wong <[email protected]> wrote:

> 
> 
> 
> On 28 June 2015 at 15:04, Pat Wong <[email protected]> wrote:
> Deke 
> 
> Thanks for the reply, yes ive seen these Videos before and whilst i believe 
> the techniques in the video will work for most situations where the camera is 
> locked off, Things tend to get a bit more complicated when the camera plate 
> is moving and there are subjects in the shot,
> 
> 
> On 28 June 2015 at 01:26, Deke Kincaid <[email protected]> wrote:
> Pat, to answer your question the workflow with the existing nuke tools 
> either: 
> 1. breaking the latlong images into six packs with the sphericalTransform 
> node. Fix in cubic, then reconvert back to latlong.
> 
> 
> How do you deal with this if an object that needs to be removed travels along 
> multiple packs. Logically i assume you'd just tile up the six packs and deal 
> with it that way, But the image dimensions might end up potentially large...
> 
> 
> Laying the cube maps into a horizontal cross pattern like this eg. 
> http://wiki.polycount.com/wiki/Cube_map
> 
> objects crossing the seams from the UP cube diagonally to FRONT cube i can 
> see that as a potential  problem, how would one deal with that?
> 
> 
> 
>  
> 
> 
>  
>  
> 
> 
> 2.mapping the latlong onto a sphere and then projecting images onto the 
> sphere to patch and fix images.
> 
> 
> 
> Would  projecting onto a sphere would result in a lack of parrallax? so im 
> guessing you might need geo of the scene ideally?
> 
> 
> Pat
> 
> 
> 
>  
> Frank Reuter did some videos 4 years ago showing some of these workflows on 
> making HDR's which is exactly the same as working with VR.  Also he has some 
> nice stitching with cards videos in there.
> 
> https://vimeo.com/album/2567386
> 
> On Saturday, June 27, 2015, Pat Wong <[email protected]> wrote:
> Any of you Nukers on the forums like to share any tips or experiences for 
> Comping in LatLong Space.
> 
>  Its a new exciting arena. One in which i just delving into...Ive just seen 
> the google stories specifically 'HELP' with its comped 3d Lizard and it 
> looked great.
> 
> 
> I've also seen the recent Nuke tech demo at NAB this year and can see the 
> people are still developing techniques to how to actually use the more 
> traditional 2D comp tricks in this LatLong distorted space.
> 
> 
> 1. Rig removals are a case in point, is the best way to do to tackle these 
> spherical transform the rig into the less distrorted central region and then 
> rotate it back to its orginal or is their a better way. If this region needs 
> to be tracked.. how do you approach this? 
> 
> 
> 2. stabilizing plates. whats the best approach? 
> 
> 3. The LatLong blue that the Foundry demo'd in the Blink Script was 
> interesting as the blur strength was obviously stronger at different parts of 
> of the distorted plate. 
> 
> This lead me to think could we somehow use a sphercial distortion map for all 
> our nodes or a utilize a STMAP UV map in some way.
> 
> 
> It would be good to hear from other peoples experience with comping in his 
> LatLong Space , on what TO DO and NOT TO DO.
> 
> 
> 
> cheers
> 
> pat wong
> 
> 
> 
> 
> -- 
> -----
> Deke Kincaid
> M&E OEM Development Manager
> The Foundry
> Mobile: (310) 883 4313
> Tel: (310) 399 4555 - Fax: (310) 450 4516
> 
> _______________________________________________
> Nuke-users mailing list
> [email protected], http://forums.thefoundry.co.uk/
> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
> 
> 
> _______________________________________________
> Nuke-users mailing list
> [email protected], http://forums.thefoundry.co.uk/
> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users

_______________________________________________
Nuke-users mailing list
[email protected], http://forums.thefoundry.co.uk/
http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users

Reply via email to