Sequencer export bit depth - for Scene Depth World Units

I want to export an EXR with the Scene Depth World Units custom render pass from Sequencer. I’m assuming that the purpose of this pass is to encode into each pixel a value for how far away that pixel is from the camera. For example, a pixel value of 500 would mean that pixel was located 500 units away from the camera. This is the effect that I need.

However, no matter what settings I use, the output seems to only be 8-bit, meaning that the values are clamped between 0-255, essentially making the pass totally useless.

I have checked ‘Capture Frames in HDR’ which does mean the resultant image sequence is EXR rather than PNG, but it’s 8-bit so it’s not actually HDR at all. I have tried all the available ‘Capture Gamuts’ and they all produce the same result.

How can I get an actual HDR (16 or 32-bit) image out of the renders from Sequencer?

When spitting out custom render passes we just use whatever format the render target is as the bitdepth for the resulting EXR, which should mean that you get half-precision EXRs by default. If that’s genuinely not happening, is it possible that the render target has fallen back to an RGBA8 format for some reason?

Is this only affecting the scene depth world units pass, or does it happen with other passes as well?

I feel like I’m missing something fundamental here – how do you set the bit depth of the render target?

Yes, every pass I’m exporting is coming out 8-bit even though I ticked HDR.

How are you determining that the output EXRs are 8-bit? OpenEXR only supports 16 and 32 bit floats to my knowledge. Could you check what the output of exrheader is with the rendered EXRs?

I am a fool.

I was using the Info panel in Photoshop, running my mouse over various pixels to see their values. Not only did I forget to set the panel to display the correct bit depth (was incorrectly assuming it would alter that automatically based on the bit depth of the file), I also forget to check whether the file was the correct bit depth by simply looking at Image->Mode.

They are indeed the correct bit depth and everything is working correctly. I’d delete this thread but for the possibility that someone else is as silly as me and makes this same mistake.