Match dynamic range of render to dynamic range of camera tracked footage?

Question

Most tutorials on camera tracking tell you to change the View Transform from Filmic to Standard. In the documentation it says Standard is "Often used for non-photorealistic results or video editing where a specific look is already baked into the input video."

This sounds good for the camera-tracked video footage but not for any 3D elements you then add into the scene afterward.

If I understand Andrew Price's video "The Secret Ingredient to Photorealism" then I should be trying to get the dynamic range and the middle grey value of my render to match that of my tracked video footage.

How do I go about doing this? It sounds like changing from Filmic to Standard just ignores the problem?

I suspect the answer involves a year of study of the nature of lighting! I've already looked at this Blender SE answer from @troy_s and both Creative Shrimp's filmic tutorial and Blender Guru's one (both of these cover filmic when it was still an externally developed plugin).

But I just got lost in the details and didn't come up a clear set of steps I should take for matching renders to tracked footage.

Is this something that's probably beyond the hobbyist or is there something slightly more clever that I can do than just switching to Standard?

Answer 1

mqbaka was right about exporting and combining them in a different file or scene. From my experience in VFX, this is the best way I've learned to go about it:

1. Have your scene tracked, modeled, lit, etc. to match your background footage, with the desired color space for the 3D scene alone. Filmic, Filmic log, whatever works best. Don't worry if it affects the color of the background footage during playback, because next you will...
2. Render your 3D scene without the background footage, just the 3D scene with transparency, to an image sequence or video file. An EXR can keep it less lossy and preserve the raw values.
3. Create a new scene in your file (or make a separate file entirely). Import your background footage and your render as nodes in the compositor. Keep the scene color space in Standard (doing the default VFX or Video Editing files does this for you). In the compositor, if you use images or image sequences, you can set the color space for each one (movie node doesn't have this, but the image sequence node has an option for movie files anyway) - the default Linear works for renders, but tweak as needed so colors match what you expect.
4. Do color matching so the render's color range matches your background. Determine what in your render should be "white" and "black" compared to your background image/video's color space. I recommend using the color balance node with the mode "Offset/Power/Slope (ASC-CDL)" as it doesn't assume your image's color space. There are some nodes that assume a 0-1 color space (some of the blend modes of RGB mix node, for example), so keep that in mind when compositing.
5. Merge your layers together via Alpha Over, setting Premultiply when necessary.

Blender seems to encourage to keep it all in at least separate scenes, if not separate files (similar logic when integrating a 3D scene into the video editor). So keep your main 3D scene in Filmic or whatever you need the render's space to be in, and then composite it separately so you can keep the color space.

Edit: an example was requested, examples files courtesy of Compositing Academy. The color grading is already in the files, but I added the Color Balance node for reference. I grouped/hid the non-relevant nodes.