32bit & 16bit workflows as compared to Photoshop

[CJI]

Hi all,

 

New to the forum here. I am a professional 3d artist, my usual software being 3ds Max, Corona Renderer, and Photoshop for still images. PS is the workhorse in my work whether it be for image manipulations, or texture creation and can run the gamut from straight forward colour correction on rendered images, to compositing 3D images in to photography. A friend of mine let me know about Affinity and i was very excited to find something that could potentially give Adobe a run for their money and break the monopoly and sway they currently have especially with their licencing structure.

I am having some trouble developing a workflow and pipleline in affinity and was hoping somebody here who knows affinity better could shed some light on this for me. My usual workflow for strict photoreal rendering is to do the main 32bit CC in my virtual frame buffer (established workflow for us, for those not in the know of 3D consider it a RAW editor), I would then take it to photoshop at 16bit, convert to smart layer and apply some additional work if needed using Camera RAW using Clarity, Shadows and highlight etc to pull some extra detail and contrast without crushing the blacks or losing the whites too much. I can then do additional layer work and painting/masking above this smart layer in the stack. If there is iterations and updates to the base render its just a case of replacing the smart layer which reuses all my previous work so that everything is consistent. This non destructive workflow is pretty important and avoids as much rework as possible.

Now to the problem. I tried to adopt something in affinity but just cant get anything to work the way I expect it to. My main niggle is in the way the colours are working. I like the Develop persona as 32bit seems to work more predictably as i would expect. However you can only save a preset here before applying the developed settings, no quick intuitive way to achieve the simple non destructive settings, plus the sliders don't work the way i would expect either in that they max out way before they would in Camera RAW. Its convoluted but i could maybe make that system work. But wait Affinity is 32 bit i think, so i can just use adjustment layers and live filters in the photo persona to do the same. Again convoluted but would preserve the non destructive nature i need. However even though its 32 bit the colours behave drastically different using individual adjustment layers compared to develop persona. Clipping and washing out and just don't work at all the way I need them too. I also tried a like for like comparison in 16bit but the colours clip far earlier than they do in PS.

Am I doing something wrong here or is it the way Affinity works for now? Please take this as constructive, Im not in any way criticising the software just highlighting shortcomings as it pertains to me. I love Affinity and think its fantastic software and wanted desperately to be able to shunt adobe to the side after years stuck in their ecosystem. I pine for the days when it wasn't subscription based and the software was released in far less buggy states. However as it stands I cant compromise my work for my want to replace it.

I made sure my profiles were set to 32bit or 16bit etc so as far as i know it is just the way it works. Really hoping to be proved wrong and its just my inexperience with the way affinity works.

Any help would be hugely appreciated.

[Dan C]

Hi @CJI,

Welcome to the forums

The first thing that comes to mind is that Affinity apps work differently when it comes to Colour Management than other creative apps. The Affinity apps perform what is called document-to-screen colour profile conversion, which can cause confusion for customers used to PS.

Can you please confirm for me, do you have a Colour Profile set for your monitor within your OS settings?

Thanks in advance!

[CJI]

Hi @Dan C

Thank you for taking the time to come back to me here.

Just the standard sRGB colour profile, I've never set a dedicated profile as never needed to.

Thanks

[lepr]

1 hour ago, Dan C said:

The first thing that comes to mind is that Affinity apps work differently when it comes to Colour Management than other creative apps. The Affinity apps perform what is called document-to-screen colour profile conversion, which can cause confusion for customers used to PS.

Photoshop does do document to screen colour management. Many creative apps (and others such as some Web browsers) do. Only a minority of my  apps do not do that.

[James Ritson]

Hi @CJI, I'll do my best to address the issues you've raised. I do a lot of 3D retouching work from Blender and 3ds Max/V-ray (EXR output) so will hopefully be able to give you a few pointers here.

First of all, I assume you're using OpenEXR or Radiance HDR formats in 32-bit? 32-bit in Photo is obviously a little bit different as it's not just linear but also unbounded, whereas Photo's 16-bit and 8-bit formats are gamma corrected and bounded. This probably means there are two main points to address: tone mapping and compositing.

How would you usually achieve tone mapping in Photoshop? In Photo, one option is to use the Tone Mapping Persona (not Develop), which will give you methods of mapping unbounded values to within the 0-1 range. You can also use OpenColorIO transforms—for example, with Blender, you can apply the Filmic view transform and looks. I did a video on that a couple of months ago: 

You can also try various adjustments for a more manual approach—the Exposure slider to shift extreme highlights down, for example, then Curves and Levels with a gamma adjustment.

Quote

However even though its 32 bit the colours behave drastically different using individual adjustment layers compared to develop persona. Clipping and washing out and just don't work at all the way I need them too. I also tried a like for like comparison in 16bit but the colours clip far earlier than they do in PS.

This brings me onto compositing—everything operates in linear space (scene linear) within 32-bit, then you have a gamma corrected view transform applied afterwards. It does mean that adjustments in particular may behave differently or seem more "sensitive".

Photo allows you to use pretty much the entire roster of tools, adjustments and filters (with the exception of median operators) in 32-bit, but there are a few caveats. Most adjustments should avoid clipping or bounding values, but pre-tone mapping I would stick to Exposure, Curves, White Balance, Channel Mixer, Levels etc. Brightness & Contrast will only operate on a 0-1 value range, but won't clip any unbounded values, and Curves will operate on the 0-1 range by default but you can change the minimum and maximum input values—so if you wanted to only manipulate bright values above 1, for example, you can set minimum to 1 and maximum to 100 (or whatever the brightest value is in your document).

Adjustments like HSL, Selective Colour and others that tend to focus on colour manipulation are best saved post-tone mapping, where your pixel values will be in the 0-1 range. If you use OCIO adjustment layers (or check out my Blender Filmic macros which do away with the OCIO dependency) you can add these adjustments above the tone mapping layers or group. If you want to use the Tone Mapping Persona, I'd advise you to do Layer>Merge Visible and then tone map the merged pixel layer, then put your adjustments above this.

Hopefully by following the above advice, you'll avoid the clipping and washing out that you describe. I suspect you may not have tone mapped your image and are trying to use adjustments like HSL, Selective Colour on the linear values? Converting to 16-bit at this point will not help the issue, since unbounded values outside 0-1 will be clipped. You need to tone map first using methods described above, then you can manipulate colour freely.

That said, as I've covered above, there are certain colour manipulations you can do on the linear values pre-tone map. Channel Mixer, for example, won't clip values, nor will White Balance. I also do a lot of stacked astrophotography editing in 32-bit linear, and sticking a White Balance adjustment before tone stretching is a really powerful way of neutralising background colour casts. It's useful with render compositing too since you can completely shift the temperature and tint without introducing artefacting.

One final caveat, then—have you configured OpenColorIO at all with your Photo setup? This throws people off, because when you do have it configured, opening an EXR or HDR document will default to the OCIO transform method (the final document to screen conversion) rather than ICC-based. This is great if you intend to export back to EXR/HDR and simply want to do some retouching work, but if you plan to export from Photo to a standard delivery format like TIFF/PNG/JPEG etc you need to be using ICC Display Transform for an accurate representation. You can configure this on the 32-bit Preview Panel (View>Studio>32-bit Preview).

To follow on from this, you mentioned profiles at the bottom of your post—I think you might be referring to document profiles etc? Don't get too hung up on this—in linear unbounded, the colour profile is somewhat arbitrary, and is only used with ICC Display Transform to convert and bound the linear values you're working with into gamma corrected values. With Unmanaged or OpenColorIO view options, this profile does not matter. If you're aiming for web delivery, stick to sRGB (Linear) with ICC Display Transform set and everything will be fine!

Apologies for the small essay, hope you find all the above useful.

[CJI]

Hi @James Ritson,

 

Thank you for such a detailed response. I really appreciate the effort here.

Yes i render out to EXR. To explain my general workflow i saved out all images from 3ds max as true linear images without tonemapping. I would then open them in Corona image editor and do my tonemapping there to bring the image down to better values closer to 0-1 (i could do this in the frame buffer within max if is so chose but this way provides maximum flexibility with renders for post production). I then save this image as 16bit exr with the tonemapping baked in. From here i bring it to photoshop and do some small bits and pieces like clarity, and playing with contrast, shadows and highlights in camera raw, If i needed to hand paint anything this is when i would do it before saving out as a jpg/png etc. The main benefit to this workflow is that is that changes to the base render are very simple, quick and consistent to perform, i can go back at any point and its a very fast process. Speed and flexibility being key here.

37 minutes ago, James Ritson said:

Photo allows you to use pretty much the entire roster of tools, adjustments and filters (with the exception of median operators) in 32-bit, but there are a few caveats. Most adjustments should avoid clipping or bounding values, but pre-tone mapping I would stick to Exposure, Curves, White Balance, Channel Mixer, Levels etc. Brightness & Contrast will only operate on a 0-1 value range, but won't clip any unbounded values, and Curves will operate on the 0-1 range by default but you can change the minimum and maximum input values—so if you wanted to only manipulate bright values above 1, for example, you can set minimum to 1 and maximum to 100 (or whatever the brightest value is in your document).

Adjustments like HSL, Selective Colour and others that tend to focus on colour manipulation are best saved post-tone mapping, where your pixel values will be in the 0-1 range. If you use OCIO adjustment layers (or check out my Blender Filmic macros which do away with the OCIO dependency) you can add these adjustments above the tone mapping layers or group. If you want to use the Tone Mapping Persona, I'd advise you to do Layer>Merge Visible and then tone map the merged pixel layer, then put your adjustments above this.

Hopefully by following the above advice, you'll avoid the clipping and washing out that you describe. I suspect you may not have tone mapped your image and are trying to use adjustments like HSL, Selective Colour on the linear values? Converting to 16-bit at this point will not help the issue, since unbounded values outside 0-1 will be clipped. You need to tone map first using methods described above, then you can manipulate colour freely.

I do appreciate what you are saying. Photoshop is not fantastic in 32bit, hence why i tonemap outside and bring in a 16bit bound image rather than working exclusively in photoshop. Having worked in Linear workspace most of my career the pitfalls of it are burned in to my brain now  To describe what i am seeing better, i am bringing the same 16bit image in to both Photo and PS and performing the same adjustments side by side and PS seems to just have more range in it which is what i am finding so weird. I tried doing all my 32bit tonemapping in photo just as test and i can certainly say it handles this considerably better than PS but doesn't beat my current way of tonemapping. Blender has certainly come on leaps and bounds and is a real contender, autodesk etc must be worried as they are really disrupting things. The filmic macros are essentially clamping highlights, applying different contrast curves, tints, saturations, white balance and applying LUTs is that correct?

I haven't configured anything in photo in terms of OpencolourIO, i was just trying a like for like comparison. Without going too much in depth as to why they are different./ This was the reason i asked here, as knew there would be some differences in the way the software handles colour and it would likely take me a while to find these on my own.

Thanks again for sharing your expertise here, its really appreciated. I will continue experimenting with it and see if I can make it work.