I try to double the height of a render with rendering one part of the pix and then tilt the camera and render more sky and then put 'em together in PS. Was a nice plan but just not working out well as the 2nd render's bottom part and the 1st render's top part just not connecting well. I guess due to the lens torsion of the camera? Not sure...
Does anybody know a trick how this can be done? (Guess best trick would be to get the full version of TG3 :D)
You can do this, but you need (quite?) a bit of overlap to make it work.
I recommend to use something like Microsoft ICE to stitch the renders together.
There are plenty of other stitching apps out there though, like PTGui.
Thx TU. Downloading ICE ;)
4-5 years ago when i had only the free version i used that method very often.
I used mostly Microsoft ICE. Sometimes it works very easy sometimes hard.
Try it first with that or another program.
If it doesn't work as you want, render more images.
I would take 3 in your example at least. Bottom,middle and above with a good overlap.
You could take more renders and crop it after.
All depends on your image and how big you want it to be.
Because of the different angles the finished image doesn't look like the one you would take from one camera.
You can play with the camera perspective if you need too.
But in general it works quite nice.
There are 2 ways in Photoshop. File > Automate > Photomerge (Perspective) or load each image into its own layer and Edit > Auto Align layers (Perspective). You will need a little bit of overlap, with enough detail to let PS align the images.
Thx Kadri and Bigben. I will try ICE now and for sure will try the PS trick since never heard of it before ;) :D
Is there a particular reason that you just couldn't increase your FOV and render at a higher resolution?
Quote from: Upon Infinity on August 11, 2014, 10:40:20 AM
Is there a particular reason that you just couldn't increase your FOV and render at a higher resolution?
Unfortunately yes and called TG3 free :D
What is FOV btw :-\
http://en.wikipedia.org/wiki/Field_of_view
Thx a bunch Kadri ;)
In general, the more overlap the better. But of course like with everything, the trick is to find the least amount of overlap with the best results. Its true with photo panos, and more true with renders. With photos I like to overlap by about half or a little less. But that would be a lot of rendering in TG. Photos are of course a lot less time consuming than rendering, as everyone one here knows. :)
Since you are using free ware, I doubt that you will want to pay for a pano soft, but "Auto Pano Pro" is the soft I like. And Even with a paid full version of TG, there are still lots of creative reasons you would want to use stitching software. There are a lot of classic photography techniques you can apply to TG, just as you would to an image created with a real DSLR.
Its probably true of lots of 3D soft, but I find it really clear and understandable with TG. I have more problems with other 3D soft... In terms of treating renders as photos or using photography methods in or on a render.
For renders it's actually easier than photos and you shouldn't need much overlap at all. Photoshop does a good job at aligning images if the images are sequential (each one overlaps the next) and doesn't need much overlap to work effectively. I've used this as a standard workflow for stitching electron micrographs. The other nice thing with Photoshop is it's blender which many people use for blending their aligned imaged images from pano software.
Hey Ben,
With photos, the more overlap the more "control points". The more control points the less distortion. It helps a ton with ghosting too.
With renders, there is no distortion (or much less) and no "ghosting" anyway . So perhaps that is why it is 'easier'?
But take for an example a fish eye pano, with little overlap, there will be problems when stitching.
And a pano with moving action in the scene will have tons of ghosting, even just from wind... But again, no such problem in TG :-[
Asked and answered ;) With renders there's no distortion, no parallax error and no movement. In line with the GIGO principle then, your providing high quality data to an alignment algorithm, so it will produce an accurate result with a small amount of overlap. Photoshop's alignment doesn't do spherical panoramas, that's a different kettle of fish which wasn't the question in this case.