I've read some articles about taking photos of glasses and most of them involved three lights. I just have two lights so far, waiting for the new Yongnuo flash before getting a third. However, it should be possible to get a decent photo with just two, right?

The task: Take a photo of 2-3 Cognac bottles with some drinking glasses, grapes, and some leaves on an old fashioned wooden desk. Similar to this:

This is just a rough example of what I have in mind. I have no clue how many light sources are used this image nor do I want to create an exact reproduction of this photo. See my above description of my actual goal.

still life

A friend asked for this, he collects Cognac and it's a pretty interesting learning task as I already figured out.

Honestly, I have no clue how to get the light right in this scene the best without a lot of trial and error. I have a 580EX II and a 430EX II, a 60 and a 90cm umbrella and a snoot with a grid plus some gels. Also a round ~60cm reflector and a large ~2m² Lastolite background that could be used as Gobo as well.

Any advice on how to get a nice smooth lighting without hard reflections on the glasses? How would you arrange the flashes the best?

  • \$\begingroup\$ See updated text. I don't know why people think I want that oil painted look? \$\endgroup\$
    – floriank
    Apr 10, 2017 at 11:43
  • \$\begingroup\$ Updated it and the text hopefully makes it clear that this is really just an example of the whole setup. \$\endgroup\$
    – floriank
    Apr 10, 2017 at 12:20
  • \$\begingroup\$ Two lights is enough... I used to have a bunch of white or black cards in the studio to reflect or block the light as desired. Also some translucent paper, aluminium foil, etc... you may also simulate windows with cut-outs, etc... Put yourself in front of the subject, move the lights, place the cards, see what is the effect on the reflections/blocking of light. \$\endgroup\$
    – roetnig
    Apr 10, 2017 at 15:08

4 Answers 4


You can get to the desired result with post processing of pictures taken at different lighting conditions. You must make sure the camera is mounted on a tripod, and that nothing in the scene moves. You take the pictures such that the light source illuminates the scene from different angles so that the reflection is at different places at each picture. For every spot in the scene there should always be at least one picture where there is no reflection at that spot.

You then align all the pictures (even pictures taken on a tripod will drift slightly in alignment), and transform them to linear colorspace. The next task is to add up the pictures using masks that block out the reflecting parts of each picture. Here you need to be careful with introducing artifacts due to a contribution from one part suddenly missing. You can do this by considering the regions around the reflected areas, you can find an approximate linear combination of the other pictures that is a close approximation to these areas.

Suppose that the region slightly away from a reflection in picture 1 is well represented by 0.4 X + 0.6 Y where X and Y represent the gray values of the pixels from pictures 2 and 3. Then you make a hard mask that covers the reflected area plus a bit more, say 12 pixels around the reflected area. Then you apply a Gaussian blur of a few pixels radius to the mask, which makes the mask gradually transparent in a small region away from the 12 pixels distance. You then replace picture 1 by

(picture 1) * mask + (0.4 picture 2 + 0.6 picture 3) * (1- mask)

This has the effect of seamlessly replacing the reflected area from picture 1 by the closest approximation from the other pictures. After having erased all the reflections, you can add up all the pictures and convert back to sRGB. Because we're working in linear colorspace, the linear combinations we take could have been obtained by applying some (possibly very complex) way to illuminate the scene (except when negative coefficients appear), the result should therefore look natural.

  • \$\begingroup\$ That's a great explanation. I'm curious if you know of a link to a tutorial/demo/story illustrating the concepts. \$\endgroup\$
    – user50888
    Apr 7, 2017 at 21:37
  • 1
    \$\begingroup\$ @benrudgers I've written quite a few ImageJ macros that help with doing such work. I'm not sure there are tutorials for this specific task. My experience with working with software like ImageJ where you have to do a lot of programming yourself is that you become quite proficient at doing post processing work like this. \$\endgroup\$ Apr 9, 2017 at 20:50
  • \$\begingroup\$ That makes sense. Writing code to process images probably falls into the realm of 'time and effort to a degree most people will not complete' that I mentioned in my answer. Not sure I'm quite to a set of use cases for that...there's probably some Lua scripting of Darktable first. \$\endgroup\$
    – user50888
    Apr 9, 2017 at 22:52
  • \$\begingroup\$ While I certainly see this as a possibility it is pretty modern ;) I don't mind doing it this way, but my goal is to understand how to create the lighting conditions I want on site, not in post production using photoshop. :) \$\endgroup\$
    – floriank
    Apr 10, 2017 at 11:42
  • \$\begingroup\$ @burzum You can indeed try to avoid doing much post processing. However, in the approach described in my answer, you would still need to work with a many different lighting conditions, so it's not going to be a matter of using post processing to fix deficient pictures due to bad lighting. You can actually use photoshop to fix such deficient pictures, but that's not what one should aim to do, as it then becomes more like painting an imagined scene.... \$\endgroup\$ Apr 10, 2017 at 20:08

That is not a good example, as @ben rudgers already comented, it could be a painting (or an HDR processed image).

But let us analize the reflections. It could be a one light ambient setup and in any case one for the vingete of the background.

So do that.

Having only two flashes DOES NOT MEAN you only have two lightsources!

Use white foamboards, use mirrors, use vegetal paper as a big diffusive softbox, make a fake window, put all your still life inside a big box, usw a "flash painting" technique, combine with natural light...

There are many things to work to model light.

My recomendation to start with your "lot of trial and error" (forget the "without" word)

Google tutorials on ONE light setup. "still life one light" and then build from that. Now your second flash is free to keep experimenting!


Reflection of what appears to be a window is a big feature of this picture. The easiest way to emulate it in a photograph could be to go to a room that HAS a window! I can also see possibilities in shooting against a black screen and adding the background later.

This picture is also very much about the choice of colours in the subject.

It's getting more and more 'not about the gear' every day! I manage a lot of photography with a handful of cheap LED light sources (see picture for the sort of thing I mean) and a variety of (often improvised) reflectors. I might use the on-camera flash for fill-in. Colour balance can be easily corrected digitally.

enter image description here


[Edit: Original image presented in question to provide context for this answer] enter image description here]

[Original Answer]

The image appears to be a painting or a digital artifact designed to appear as one. This suggests:

  1. That a lot of work went into creating the image and therefore one property of a photograph that is similar is likely to be a lot of time and effort to a degree most people will not complete (just as most people do not create paintings or similar digital artifacts).

  2. As a piece of art, there is no direct evidence that such an image can be recreated as a photograph directly from a single scene in a single shot. For example are the reflections in the glass and decanter geometrically consistent? Considering the decanter's handle looks a bit 'off', I suspect that the image creator used their artistic license.

Which is why I don't think there is a way to get this sort of lighting without a lot of trial and error and probably it will require quite a bit of post processing as well.

  • 4
    \$\begingroup\$ I think it's quite obvious that the example is a painting, and it's equally obvious that the OP doesn't want a photograph that looks like a painting. The painting is an example of the sort of composition he wants (glass objects plus some fruit), and how he wants said glass objects to look (like they're reflecting a natural light source, not a camera flash or studio light). \$\endgroup\$
    – Martha
    Apr 7, 2017 at 23:37
  • \$\begingroup\$ @Martha The dynamic range of the scene ranges from the window (EV14 cloudy bright) to an indoor interior (EV 5). Photographically, with 12 stops dynamic range, the shadows are approximately EV -6 (not pure black). The artist has compressed roughly ~20 'fstops' of human vision into the ~7 'fstops' possible with pigment. Because an ordinary 'top of the line monitor' is about ~10 fstops (10 bits) a lot of dynamic range compression will still be required to represent the scene. A professional DSLR has 14 fstops/bits and a high end medium format digital camera might have 16 fstops/bits. \$\endgroup\$
    – user50888
    Apr 8, 2017 at 17:47

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.