I've always seen people say that, as a general rule, things such as physic simulations, etc should be done on a real world scale.
I almost always model in real world scale but any time I try to run a sim things tend to go completely bonkers and I struggle to get settings dialed in. I know Kent is working at a much larger scale throughout most of the Human course. Would it be advisable to scale models up, even just temporarily, to run simulations?
I have not heard that said in regard to physics sims. But I would agree.
Modeling, yes. Most definitely try to model to real world scale. However, like you said, Kent did model much larger in HUMAN.
Side note: It does depend on the future use of the model, we tend to model for the project we are working on and real world scale may not be important in this project.
However, the model may have a use in a future project that does need to conform to real world scale. Why remodel when you can reuse?
In the example of HUMAN, I would suggest at the time of course production Kent modeled for the sole purpose of educational content. Of course the final model can always be scaled to the correct size if needed.
Back to the question...
It may be that the scale is not applied, not just on the object with the physics simulation, but also any particle or collision objects for example.
I always find the default for settings like stiffness, damping, collision margin, etc., OK (with some minor tweaking) for larger objects/scenes. For smaller, these will always need much more attention and adjustment.
There are so many settings in the physics properties, it does take some time to dial in what works best.
Following a physics tutorial doesn't always produce the same result, if the model/scene is bigger or smaller, the settings will need to be adjusted to compensate.
There may also be settings elsewhere in Blender that are different. In the scene properties the units of measurement may be different, or the gravity is different to the physics property.
To answer the question:
I would say no, because some settings may change the simulation quite considerably. An example of a simple collision system of a ball dropping onto a surface and rolling down a slope, the friction and damping settings can make a difference to how heavy the ball looks, for a larger ball it may look ok, but scaling down to a smaller ball may look odd.
Physics settings IMO, are a per object/scene dilemma.
CORE is due for release s̶o̶o̶n̶ today, check out the all new Fundamentals of Physics course for may cool examples and many settings discussed.
I would agree, I always model to scale and if things are not matching up with a video I'm following, I'll just dial down my own numbers. It's just good practice to model to scale, everything from lights power and values will work correctly and in time you'll develop a cohesion in your mind and know what to dial in by feel. So nah, don't scale up, just dial the numbers corresponding of your scene I'd say.
It really depends on what scale you're talking about. For example: rigid body and dices. Because of precision rounding limitations you can't set anything below 1 gram in weight. Although you can do any milligrams over 1 gram.
Now with cloth sims mass is often times overlooked, but very important. Defining the resting shape is also very important. I will run the sim until a find a resting position that I like. Sometimes I'll have to get close apply then run sim again to get a good starting point(resting position). 2 other common problems are scale transform and flipped normals.
I should have clarified, by real world scale I mean a human figure as real world scale. Also, I'm asking in general after watching the video rather than following along - when I experimented with my own Human project in a much larger scale the sim behaved better. At the time I thought it was due to the below mention mass setting but not entirely.
In my work I've always got the scale applied to all objects (after all these years I've learned it should always be the first thing I check), check normals, merge overlaps, find strays, all the usual suspects. Playing with mass is definitely something I attempt but I admit to not understanding it. .3kg per vertex? I sew iRL and I understand fabric weight (like, 300gsm denim = ~1 sq meter of denim etc) but I've yet to grasp correlation between RL weight and mass in blender (I admit, the only real understanding I have of weight vs mass is what I've learned in EVE and Kerbal Space Project having inadvertently dissociated my school years).
I do try to just experiment but when so much goes crazy all at once it's hard to even tell where to start. As it goes with other issues I've come across that seem to arise from working at RL scale (specifically, hair nodes) most of the examples I see are done at larger scales with less problems. I'm thinking I'm going to have to scale things up in order to get consistent results and hope that scaling back down goes gracefully.
The fact things consistently go wacky for me suggests that it's user error though so I do appreciate all the troubleshooting advice!
It's probably not user error. In my experience It's the 1.5CM(0.015M) distance under object collision and the vertex mass. Both of which work better when you scale up a human character, but work great with blankets and curtains. As for the vertex mass, I just google the cloth type and material to find it's mass. Like "Weight of a Cotton Shirt". Most of the time they include grams(kilograms) along with the ounces.
I totally agree with the philosophy you're describing coyohti. It's been a minute since I recorded HUMAN but I'm pretty sure the non-real-world scale aspect is something I wish I'd done differently. Admittedly I've dismissed real world scale too much in my journey. It's not a required technique but I've grown to appreciate the utility it provides with things like simulation parameters.
@theluthier - In the case of cloth sim I'm feeling like you are on track as far as working in a larger scale! I've been doing a bunch of experiments at real world scale and it's still just very, very fussy. I think working at a larger scale somehow gives more wiggle-room somehow.
For instance, dillenbata3 mentions the .0015m collision distance. This works fine on a larger object but scaling it down to a reasonable distance on a real world scale object (say, even 5mm) seems to confuse the simulation. Even if the cloth object starts far outside the collision distance something wacky happens and I get pulsating spikes of verts.
I've been testing with a clothing object intended to be a jersey knit tank top. Even at real world scale I found that pushing the vertex mass up to 1kg or even over helps the mesh fall in a way more in keeping with how it would looking iRL. However, there are still issues. The lower curve of the armscye always wants to bunch up. I noticed the same thing in Kent's hoodie example but I think, in that case, it was because the armscye was not deep enough to accommodate the model's massive guns. I've added more positive ease to the area under the arms and it (understandably really) just exacerbates the issue.
I think if there is one tool from Marvelous Designer that I'd like to see in Blender it's that thing where you can sort of pluck at the "fabric" to get the sim to rework certain areas. I can dream! :D
coyohti I'm not positive, but wouldn't the cloth brush in sculpt mode be similar to the Marvelous Designers tool. Under the hood the cloth brush/tool uses the cloth sim. Sorry, I don't know much about sculpting. I'm pretty terrible at it. Even when I've barrowed my brother-in-laws drawing tablet.