I do use the matchbox from 3de which is excellent. However, we are trying to build a pipeline that is robust and having two exports seems too much for tracking. I know. It’s a fuck-on.
This is awesome. I’ve never tried the 3DE matchbox. Gotta check it out!
It’s also only for 3DE…so the issue wouldn’t be solved for STMaps from other applications. @cnoellert’s solve above does the trick though. While not as easy as pipe in input and STMap, it’s not far off.
Hi. Is this2021 archive?
@johnt I believe it is but I can make a 2020 version if you’d like. I should still have a version hanging around…
This is where we are at, at the moment.
I am just trying to get a solid internal workflow so that when we get lens grids we can do it ourselves.
We use Nuke for the grid analysis and then reformat the nuke clips so that we when we export ST maps we get the full bowtie extension. No lost pixels.
I like the idea of getting 3D equalizer matchbox from external tracking though.
We are using Syntheyes internally.
Trying to get 3D to render using the STmaps has been tricky. Not everyone knows how to do it.
Rich, yep we’ve had issues with redshift and stMaps but it was operator error with the map not being used at 32bit. My stance was always “3d should be giving me back the correct distorted renders at source res”. But after some discussion we came to the conclusion that it’s easier for flame to do it (only because of the 3DE matchbox workflow).
If you have the possibility to make it 2020 that would be really helpful.
@johnt, here you go. I also simplified it even further…
4 Likes
Thank you so much.
I’ve at some point used ST map lens distorts with negative values in Flame - you can work out from the values how far outside the bouding box it’s trying to read, then resize out to at least that size and use a CC node to multiply and offset the UV values by the same amount - like if you see -0.1 it needs 10% overscan on each side, so pad out from 1920x1080 to 2304x1296 then CC the UV map itself with histogram min/max in -0.1/1.1. Pretty stupid process but it did work.
I don’t think it’s possible to automate this with Matchbox sadly because you can’t set the output resolution from the shader code - only to fixed values in the XML.
Honestly I avoid ever using that “undistort past the edges” stuff even working purely in Nuke because it causes chaos for everyone, not just Flame - way too many nodes in Nuke will crop to the format rather than the bbox and lose the extra info. Renders from CG should be at an overscanned res anyway so you might as well just use that size for all work that’s done undistorted…
2 Likes
Your process to CC the UV is interesting but yeah, that’s definitely a bit much of a process. I completely agree about undistorted plates not being cropped and I’d never want to go down the route.
My interest was more in the re-distort process. When we outsource our tracking we almost always get back STMaps that hold negative values since the original plate wasn’t padded beforehand. This has lead to us un-distorting for them so we know for sure that we can comp in Flame w/o any headaches, which isn’t ideal for either side. I can appreciate why they prefer to do their own distortion but they seem to get fussy when it comes to padding beforehand. It’s resulted in a lot of back and forth and, of course, people banging on how it just works in Nuke.
Chris’ method to re-distort works a treat so that’s a bit of an eye opener. I’m looking forward to the next time we have to send tracking out and not have to go through this back and forth.
2 Likes
Where can I found the matchbox 3d Equalizer ?
I think it is exported from 3DEqualizer and it is unique to the particular shot that has been tracked
Hey @cnoellert , I was messing around with this yesterday and I noticed the distort/undistort plates were both the same resolution. Is that unique to Mocha distort/undistort stmps?
I honestly don’t know, although it would easy enough to ascertain. The mocha undistort process really doesn’t come up super often for me, but works a treat when you shoot those perky wide-angles and need screen inserts…
Copy. When applying your handy dandy helpful Perspective surface/Image Extendy thingamajiggy trick I can get a nice undistort and redistort, but the edges of the frame are ever so slightly problematic and jaggy…wondering if that will ever bite me in the tookus when patching undistorted patches into distorted plates.
Worth double checking the filter settings on your maps as well as on your surfaces to see if you can make some improvements, but otherwise it’s tough to say. I’ve seen some pretty nasty edges on stmap transforms but luckily people in our line of work tend to have a bag of tricks for when things don’t work out 100% as expected.
I can always take a gander if you wanna pop out an archive.
1 Like
Does this great script work with alembic ?
Yup, there a menu option for alembics.
2 Likes