So I’m a ways overdue now aren’t I? I put Q4 as the release date and now that’s long gone. So what’s going on with the time-lapses? Good question!
Initially I was testing the best shooting setup to capture the HDRIs as efficiently and accurately as possible, and since that time I’ve been focused on putting them together in the most accurate way possible. I need to perfect the camera curves, colors, contrast and gamma, and the short answer is, creating these super accurate HDRIs is difficult without a linear toolset.
My workflow right now consists of a rather large amount of programs, including:
- Lightroom – import and export RAW NEFs to aRGB or sRGB Tiffs
- Photomatix – import non-linear Tiffs which-are-close-to-linear-but-not-quite(who knows)… and make HDRI frames
- PTGUI – convert to lat/long format
- Adjust with Photoshop…
- Test render with Blender Octane and Maxwell Render – reveals a somewhat high contrast result…
When comparing my IBL renders to my go to reference HDRIs (Paul Debevec’s Sky Probes), my results at a similar time of day is much higher contrast, leading me to believe that at some point in my workflow, something is going wrong in terms of gamma curves, camera curves or a combination. My initial sunset HDRIs were most likely not affected as much because they had less range (the sun was obscured by cloud).
High Contrast Frame Comparison
Here’s my default mid-day-ish frame:
Compared to Debevec’s at roughly the same time:
Debevec’s looks much lower contrast, perhaps a little too low for my tastes, but the good thing with low contrast is that it is much easier to post process however you like. Once you have deep blacks and blinding whites, its harder to pull them back.
Gamma and Linear issues
The main problem I’m having with this workflow is that right out of the gate, I’m importing and exporting images with Lightroom, which doesn’t allow a linear export. So this creates a knock on effect where all the subsequent calculations are working on non-optimal starting images.
If I use something like dcraw (exports linear) I then lose all the great benefits of Lightroom, which includes chromatic aberration reduction, fringing reduction, etc.
As the contrast seemed an unsolvable problem, I rendered out the sequence anyway and then ran them through Photoshop and adjusted the gamma so that the contrast was lowered. This seems like an ok solution, but I’d much rather have a perfectly accurate solution that doesn’t require ‘eyeballing it’.
I also experimented with using -40 to -100 contrast levels within Lightroom to try and linearize the Tiffs but as I said before, its hard to know exactly what the correct setting is to try and match a linear output.
New Linear Workflow with Nuke
Starting next week or so I’ll be getting my hands on Nuke and trying a fully linear workflow. I’ll post the results here and give my thoughts on the differences, pros and cons. Really looking forward to trying it and I’m hoping that it will be a much cleaner and more accurate solution, and being able to stick to one program for everything should help a lot too.
After playing with a couple trial versions of Nuke, my options seem to be either to use Lightroom as part of the workflow (which would then not be fully linear, but have better chromatic aberration/fringing removal) or use Nuke for everything with some select scripts and plugins like J Ops for Merging to HDR and importing NEFs. Alternatively I could match my Lightroom exports to the linear Nuke exports and take it from there.
Will keep you informed!
Thanks for following – if you’d like to be notified when the free samples are (finally!) released, you can sign up to the newsletter here. Apologies for how long the samples are taking, I know I’ve said a number of times how close they are, but I just can’t release something unless its up to standards.