[HDRI] Mpumalanga Veld

Download Free HDRI - mpumalanga_veldDownload Free HDRI - mpumalanga_veld

Download Free HDRI (1024×512)
Licensed CC-BY

Want more free HDRIs? Check out my new dedicated website: HDRI Haven


This, I think, is the first HDRI that I’ve released publically that includes the full dynamic range of the sun. I came pretty close with Bergen a few months ago, but I wasn’t quite there yet.

How can you tell?

Here’s a close-up of the sun in one of my even older HDRIs, Delta, with the exposure reduced by 14 stops in Photoshop:

You can clearly see that the sun is just a solid color, and it’s not even that round. This is like what you’ll find in most HDRIs, even some you pay for, because it’s the best you can do without a filter.

Using it in Blender will give you something like this:

The shadows are not very distinct because the sun in the photo isn’t as bright as it is in real life. Of course there are some tricks you can do to fake the “correct” brightness, but this will only ever give you a white sun because the colours are still clipped in the original.

If you really want more accurate lighting, the HDRI needs to be captured in such a way that it holds all of the available light information, including the full brightness on the sun. This is really tricky, because the sun is actually really super bright.

So earlier this year, I shot Bergen while on holiday in Norway. I used an ND8 filter to reduce the brightness by 3 stops (remember that most HDRIs are shot with a dynamic range of around 15 stops, so 3 more stops is fairly significant). The result? Better…

… but, I knew deep down inside that this is still not correct. If you open it in Photoshop and reduce the exposure again (this time by 16 stops), you get this:

Again, you can see that the sun is a solid colour and completely white, but we know from wikipedia that it’s supposed to be brighter in the middle, perfectly round and slightly yellow from atmospheric scattering.

So to settle it once and for all, I managed to find an ND400 filter, which lets only 0.25% of light through it, meaning I could finally capture the sun in all of it’s ridiculously bright glory.

As proof, this new HDRI with the exposure reduced by 18 stops looks like this:

This time the sun is not a solid colour, rather it’s brighter in the middle, perfectly round, and slightly yellow.

This isn’t the first HDRI that I’ve made using this method though. I made a set of 9 (plus Bergen) as a bonus pack for the AgenZasBrother’s upcoming workshop, which you’ll also be able to find available on hdrihaven.com when it launches next year ;) </tease>

One last thing I should mention is that I actually used Blender to merge the various exposure brackets together, as the hardcore professional paid applications that are made just for this don’t actually handle super bright light sources very well. I tried a bunch of free software too like Luminance HDR and Picturenaut, but again they failed to produce the correct brightness and even added in artifacts and noise in some cases. I’ll cover exactly how I wrangled Blender to do this in another post sometime soon(ish).

  • ogbog

    How’d you merge them in Blender? I’ve been working on the same issues as you, shooting with as much as 12 stops in ND gels to try and map that fickle temptress, the sun. Great work!

    • It’s a bit complicated, but I’m planning to write about it in detail sometime soon.

      Basically you first increase the exposures of all but the brightest bracket so that they are all the same brightness (this is the tricky bit, since you have to take the camera response curve into account and do some math).

      Then using the original images, make masks for what the best-exposed areas of each bracket are (chop off the dark noisy bits and blown out whites) and mix the normalized exposures using those masks. Thus you’re left with a single image that has the full exposure range.

      • Chris Cook

        Hello Greg, the image looks great! I’ve been tackling a similar problem and just couldn’t find any hint of what the sensor response curves actually look like and how to correct them! It seems to only affect the top quarter of the histogram. How did you correct the response curve?

      • Could you share a .blend or something, I have a fully shot HDR waiting to be merged, photoshop CC’s HDR merge Pro function failed (making something that was nearly white at -12 EV go to grey at -6 EV), and I’m not quite sure what to do.

        • I think I have figured out the correct method, it will be interesting to compare results

      • Samuel Loy

        Wow… I don’t quite get it. Do you have an idea when the article will be done?

  • mmoore500

    This is an awesome hdri. I thought Bergen was good. Great work. Thank you!

    • Thanks man :) hope it’s useful.

  • Robert Coughlan

    This is fascinating. Been dabbling with HDR / spherical panos for a little while now, and have recently started learning Blender. Would be great to combine the two. Can’t wait for more detailed info on your method using Blender. I’ve not previously used ND filters, only polarizers, so that should be something new.

  • AAnderson

    A brief note to thank you for your efforts. I’m more of a designer/artist, so having someone like you tend to the technical details for those of us who shy away from them (while still providing excellent context as to the whys) is deeply appreciated. Kudos.

    • Naman

      Hi.. I want the pro lighting skies ultimate addon..
      Can you please give me in exchange i will give you the fully purchased grass essentials for free..

  • Raphael Beguin

    Ok, after reading this, i feel, i learn a different approach to create hdri. I followed similar workflow using my dslr tripod and a 50mm with only a UV filter (i will try with my fader ND), shoot 3 or more pix, then used Photoshop, to merge them, also tried Picturenaut, but feel not really convinced by the result.
    But shadows shouldn’t they be more smoother? like the top image.
    On your Bergen image the tree shadows are smoother than the other at the opposite side.
    I’m also curious about how you merge them within blender (used compositing nodes i guess) ;-)

  • Russ Sprouse

    Just bought the teaser pack, Greg. Good luck!
    ~ Russ

  • Sayan Mondal

    Awesome HDRI. Thank you man!
    -S.

    • My pleasure sir :) hope you find a use for it.

  • blackdot

    awesome, thanks a lot Greg!

  • Kevin

    Hi Greg – a fascinating post on top of your other really interesting ones in this subject.
    I’m a little confused though – are you saying that the HDRI you’ve taken is better because the sun is a better representation of the sun in real life without all the bloom and glare a normal photo (taken without filters) would be?

    Plus, by taking a series of photos with different exposures and combining the result in your software, you also get a sun which looks right but is relatively much brighter than the rest of the panoramic image.

    So, if you take a photo into Photoshop, convert it to 32bit and save as exr you are going to get the same results as the original photo since you’re just scaling the colour values up to 32 bit.

    However, if you compress/darken the colour down to say 1/10th of its original range (e.g. make everything dark) then draw in a sun (using the sun description in your post) then you would get something with a more accurate light spectrum.

    You would then have to adjust the strength of the HDRI in blender to get it to a point where the sky looks blue rather than dark blue.
    Does this make sense? Should this work in theory? (I’ve been struggling to get it to work in practice!)

    • No, the bloom and glare is still there. This is simply addressing the brightness of the sun by capturing and merging it all correctly.

      You *can* draw in the sun yourself (as Chocofur shows here), but that won’t be very accurate to real life, unless you’re Michelangelo.

      • Kevin

        Ah I see – thanks for the clarification and the link!

  • Pingback: What Makes a Good HDRI and How to Use It Correctly - Adaptive Samples()