Help? - A couple of questions

As the main question I have is about compositing, I’ve posted this here.
I am trying to do some camera tracking / compositing but incorporate a water element to it.
I am using the Blender Guru camera tracking tutorial for testing purposes and am quite happy with tracking etc.

However, I want to add water flowing down the road and into the hole but as the camera plate provides the “world” in which this takes place, the road doesn’t show through the water, so the water looks opaque rather than transparent.
Also, the water casts as shadow, due to the way fluids work, (I have lowered the baked domain so it sits on the road rather than above it).

Finally, the fluid doesn’t “stream” into the hole, even though I have used the water preset, it is very blobby and behaves more like oil, (thus I ended up using the look in the image for the fluid).

I have attached the blend file, but you would need to get the BG tracking plate files from

Hoping someone can help.

Attachments


A few observations:

(1) Obviously, your fluid is too viscous. It does behave like crude.

(2) When compositing, remember that you can do anything that you need to do in order to obtain each component that you intend to comp together. You could, for example, construct a trough for the fluid to flow down, then omit the trough (layer) from the render of the liquid-only. As long as it appears to coincide with the road in the final comp, it doesn’t need to have anything to do with the road at all. (Using linked cameras, etc. is very helpful in making sure that they do coincide and that continue to do so as you inevitably tweak.)

(2a) If one of the layers is calculated, e.g. by a fluid-sim, then you can use the calculated result … without re-calculating it … in order to do other things such as modify the opacity of points within the layer during (node based) compositing.

(3) Elements such as “the opacity of the water” can simply be a function of the compositing noodle-network. You need the simulation result mostly to know “where the water is, and what shape it is.”

(4) “Don’t over-think it. If it looks good (enough!), etc.” Decide what you want it to look like, come up with a computationally-efficient and time-efficient way to get there, then … “get there if you can, then stop and stuff the film into the metal can.”

Sorry if I come across as a total thickie, but a lot of what you said just flew straight over my head.
I understand the bit about making a duplicate scene to control the water flow, but not clear on the linked camera thing.
The water viscosity, I presume I have to noodle around with the base and exponent values for the domain, (though I’ve no idea what they actually mean, if anything logical).
Items 2 & 3 lost me completely, am I to assume to can control the water opacity /refractino etc purely by node is compositing?
What is the “calculated result”, or do you just mean the resulting mesh from the domain bake?
Or is it all about masks etc? I’ve tried a google search for some form of tutorial I can relate to, plenty on camera tracking, even on putting reflections from the scene into footage, but nothing on putting water into a scene and having it look right.

No problem. What I’d do is to set up one scene that contains things (such as cameras, significant lights, and reference geometry) that are to be common to all other scenes. Then, create additional scenes, linked to the first, that represent things that you actually want to produce to make the shot. (Or, you can use entirely separate files and still link them, but I digress.)

“First, do the roadway,” and render that roadway. (If the surface of the road and the interior of the hole are significantly different e.g. in lighting, you can also render these separately, for later compositing. Likewise the dirt field. Anything. “Break it down. Break it down …”)

“Next, do the water.” Let the water flow within a box or guide of some kind that will be “seen” by the water simulation but that will not appear on-screen (i.e. it’s in a hidden layer). Render the water. (Don’t worry about the pot-hole: you can mask that later. But also don’t bother to render stuff that will be forever blocked by the hole.

In both cases, render to MultiLayer OpenEXR files and in those files capture whatever RenderLayers you think might be useful in the next step(s).

Remember that you can link to actual geometry, such as the interior of the hole, if only to use it as a visual reference in placing the water-guides.

“Finally, compositing.” This step uses the outputs generated by the other two steps, and it is at this point purely a two-dimensional process. The water is superimposed over the road, but it is made partially transparent. You can use characteristics of the water to modulate exactly how the two (or more …) channels of information (“road,” “water” …) are combined. There are a wide variety of things that you can do at this step, and “they’re all fast and cheap,” because it’s all two-dimensional processing at this point … although it can make use of any of the data that you captured in the MultiLayer OpenEXR file format. (It can, in other words, be “3D aware.”) You can color things, for example, or tweak the colors. You can vary the opacity of things, like water. You can dynamically adjust the hue and saturation of things, like the road, to further the illusion that you are “looking through water.” You have lots of opportunity to be very creative here.

A good first step, after rendering the road in this way, is to create nothing more than a rough piece of geometry, with appropriate materials, that will represent the water more-or-less everywhere it’s eventually going to be, so that you can immediately shift ahead to the compositing step to get that “block of water” looking sufficiently “like water” in the compositor output. Use this to guide you as you go back and construct the water simulation. (It might then become your bounding-box …)

This is a multi-step workflow that will take you to a result that looks convincing [enough], and that minimizes the expensive process of “3D rendering.” The use of “MultiLayer OpenEXR” is critical because this format captures all of the digital data in high-resolution numeric form. You have “all of the numbers” that those renders produced, in addition to the pixels.

Just a question, as the road as such doesn’t exist, being a camera tracking plate, would I need to somehow “map” it to actual geometry? I am thinking I may have bitten off far more than I can chew here when it comes to my skills vs the task at hand. :o