Add transcription for: frames/MDA_TheSymbioteCreature_DownloadPirate.com_SYMBIOTE_CH009_RECAP_MOTION_TRANSFER_AND_ORGANIZING_1080p_mp4_frames.zip

transcriptions/frames/MDA_TheSymbioteCreature_DownloadPirate.com_SYMBIOTE_CH009_RECAP_MOTION_TRANSFER_AND_ORGANIZING_1080p_mp4_frames_transcription.json

@@ -0,0 +1,8 @@
+{
+  "text": " So in the previous lesson I forgot to mention that I basically added a Vellum input output. It's this node. Vellum IO. I set it to explicit. I set my path, the frame range, applied the color to the geometry output, and emerged in the ground. So now we have this. And you can see what's the result of the simulation. looks pretty good for me. There's a little bug here, the tail that's wiggling a little, which is weird, but we're gonna fix it in the next iteration. But the compression of the muscles looks nice. The bubbles are not overly inflating, so I like it. We have this result because before I spent a lot of time tweaking the parameters and trying to dial in the values that work well, so this is why in the first long iteration we have this result. but usually it takes quite a lot of iteration just to test this simulation. So, and the parameters, we have the sliding of the fail here as we planned, then yeah, we have a bunch of stuff happening here. So yeah, then next step would be applying this animation to our quad mesh to see how it looks and use that quad mesh as a source for the flip simulation. This is not the whole simulation, it's just 115 frames, so it's enough for a preview. Sometimes I split the model into pieces, just for example I isolate one tentacle just to see if it works. So I'll give you an example. So basically if you have this as a model, sometimes I would come here, put a blast node in the middle, and I would select really a piece of the geometry instead of selecting this. So click again and select some points. Do this, did it not select it. And I would simulate only this piece of geometry. So yeah, this would allow me to have faster iterations. But it depends on the dynamics. Here we'll see just that the arm is going to be just throwing itself and then this part of the body is going to follow because we don't have any tail constraining the body to the ground. It's going to be a little bit chaotic. So it works in some scenarios, it depends. But yeah, I advise you to, when you try simulations like this, I advise you to go into your Microsoftres, disable a bunch of them and then simulate again. So I prefer to just simulate separately each Microsoftre, with each Microsoftre, like separate the simulation into several steps, see if separate things work properly. Yeah, so I'll disable this. Just to recap what we did here, we had a quad mesh. We grouped a bunch of stuff together. We painted some masks. We separated bubbles. We had a remeshed object. Then we animated it with the help of some attribute triangles. We transferred the animated attributes to our tetrahedral mesh. We had a two-out triangulated. Oh yeah, we copied some attributes, animated attributes to our triangular mesh and then, then shifted our triangular mesh with tetrahedralized it, transformed it into volumetric tetrahedrons. We transferred some groups, created an IDS attribute. We also set some casting properties. The vector called their arm, which defines the direction of the contraction of our muscles. We applied some density, stretch stiffness using our relative P-Box expression in the Z-direction, which allows us to have some really good control over our density. So if you stretch stiffness, and then we have density. So these are parameters that define how heavy the mesh and the object is in separate parts of the body. Then we transferred the live attributes onto our tetrahedral mesh. We attribute blurred. so it blends better between the vertices. Then we had a master transform that defines the initial placement of our monster above the ground. So you see it's barely above the ground for the initial drop. Then we set some valent cloth constraints which define the bendiness and the stiffness of the surface triangles. Surface triangles, if you remember, were created by the teticon form of this checkbox. So it's a separate additional layer on top of the tetrahedral mesh, which allow you to make some kind of wrinkles there, but if your mesh is not dense enough, you'll barely see those wrinkles. I added it just so I could show you the default, like the initial, the meta setup for these kinds of simulation. Then we created the Valentets. We applied some tetrahedral stretch constraints with a model called nonlinear ARAP, which is as rigid as possible, which you'll choice to maintain the volume tetrahedrons as much as possible during deformation and we applied those constraints to our group which is called GRP dense tetrahedrons only not the surface. The density was multiplied by a attribute which we created earlier, the density, then we had the thickness calculated uniformly over all of all the vertices. And then same with the stretch stiffness, we multiplied it by the stretch stiffness attribute. Then for the fibers we applied that fiber constraints only on the group called grp legs which is a point attribute, a point group, and the stiffness was set to zero because it's going to be controlled inside the valum-solver. So inside the valum-solver we created a sub-solver inside of which we're importing our rest geometry which is this geometry, the alt-rest which is geometry that has the live attributes cycling through. So in the sub solver we're basically transferring the live attributes onto the dock network geometry. Then we're sampling data from the ground. So before initially we as far as you remember we tried this locomotion test here in the subs just to see if the forces are transferred. We created a FRC debug force here which is the resulting force without any multiplication. We visualized it to see if it works properly, as we intended it to work. And then once we were done with the SOP test, we transferred the same VOP inside the SOP solver, right here, and we'll lock a motion test. So it does exactly the same, but over time. So this is the magic node, which just samples the velocity and the normal from the surface. The velocity is just a directional minus one vector in the z-axis. We normalize the forces, we added them together, multiplied them by CDR. The CDR controls the normal force if it's negative or positive over time. So the CD actually that activates the force only on the tips. And yeah, so the CDR is really crucial here. And then we safely added the force to the existing force. So the force here, we take the force here and then we add our resulting force to the existing force and we apply it to the force, not velocity. This is crucial because this is the safe method of adding. We can do it on the end velocity, but there's a lot of other stuff to consider. I'm not going to cover this here, but for now it's just, know that you have to, in this case you have to add it to the force. So yeah, so this is the subsolver. The subsolver pulls the live attributes and so they exist in the simulation, updated live on each frame. Then we created this geometry wrangle that reads the constrained geometry. So, the constraint geometry checks if the constraint is a specific type, which is a wrap, the as-readed as possible constraint. If it's not, it doesn't do anything. If it is, it just takes our derarm attribute and rotates it with the quaternion attribute, which is a default attribute contained in any tetrahedral mesh with the deadstretch constraints applied to it. The rest vector is the initial alignment orientation of the tetrahedron. So with the help of this rest vector, which is updated live during simulation, we rotate our V-dir arm so it points always into the direction of the deformation of our legs or arms or tentacles. However, you may name those. Then we have this, oh well, that would have been a mistake. If I scroll like this, it's going to increase the number. So Let's be careful there. We created the Vellon constraints Microsoftr, which is creating constraints between the tips and the ground at a specific time. And then another Vellon constraint properties, which takes those constraints, processes them, only them, and checks if the color of the vertices attached to those constraints is bigger than zero. If it is, it's going to kill them off with the use of this dummy, remove property. And then we have the animate contract which takes the fiber constraints we created earlier in the sub context. And it takes the color of those vertices related to the fiber constraints and applies a stiffness to them. So if it's white, it contracts, if it's black, it just releases. It doesn't expand, it just releases. So it removes the contraction basically. So the muscle becomes relaxed. Then we created another Bellon constraint attachment, attached to geometry. After the frame 5, between the group tail ends, which is the tail of our creature, and the ground force, we set also the crucial thing here, the sliding rate, which allows those constraints to slide along the surface. They're attached, but they have some freedom. If the sliding rate was zero, they wouldn't have the possibility to move. So the yarns will pull, but the whole body would stay in place because these constraints do not allow the creature to move. And finally, the bubbles animate, which is also a well-constrained property. Microsoft, which takes all of the text-stretched constraints, which is the whole body, and then it checks if those constraints are related to vertices in the group called group bubbles. If they are, then there's a fitting happening, which is taking the color of those bubbles and fitting them in between one and the max inflate value, which is 3, so between 1 and 3. So when the bubbles are black, the inflation is 1, which is basically the default state, the uninflated bubbles, the rest bubbles, and then max inflate is 3. So the rest length of these triangles and the tetrahedrons inside is going to multiply by 3, relatively speaking, because there's connectivity involved, involved so sometimes they do not inflate to 3 because there's limits where those constraints try to stay as really as possible so they do not allow inflation above some values. If we were to use another type of constraint like for example, if we go to this to linear A-Rap and uncheck preserve volume they would be able to expand almost limitlessly but the stability of the simulation would be completely different and there might be things that would explode. So here we go, we have this setup. I'm pretty happy with the result. This setup is quite clean compared to the initial setup I had before was just a horror because it was kind of a playground for me. I was playing around with nodes but this is a really condensed and compressed and a really short version of the tutorial. I mean, not short version of the tutorial, but it's the essence of the locomotion system I developed. So yeah, let's see what we do next. I will just apply, in this tutorial, I'll just apply the motion to our quad mesh and then we'll see what we can do next. So let's pull our quad mesh. Going back here to the top, creating an object merge, this into the object. This is how I like to do the merging. And yeah, so into this object, now we have this quad mesh, but as you can see it's penetrating the ground. And this is because we have master transform to respect. So you have to respect this master transform. So we'll just have to apply the same master transform to the quad mesh before we do any transformation or motion transfer. So when you right click on the master transform, and in the action you do a create reference copy. So this is a trick where you can transfer the same positioning rotation as our tetrahedral mesh. And now we're going to drop a point deform. And the first input goes our tree mesh. And we're going to drop a time shift. The time shift, delete the channels. So it goes into second input, and the animated version of the tetrahedral mesh goes into the point deform. So yeah. And as you can see, we have our quad mesh moving. And it's cool. One thing to mention is that the output mesh, as you can see, it's tetrahedrons. So what I'm going to do here, just clean the mesh. We're the only thing I'll be leaving here, is remove all the attributes except V and CD. and CD. So I'll uncheck the color because it was for previous purposes, not for the final result. And we may just, yeah, this, we have our mesh here. And then let's see what I'm going to do here. just bake this mesh so we can have faster feedback and yeah we have tetrahedrons here so one thing to mention is you see that we have polygons so we have tetrahedrons what I prefer to do is always do the point deform from the tetrahedrons basically the motion I love when there's tetrahedrons transferring motion to a poly mesh because the tetrahedral mesh contains volume volumetric tetrahedrons and they deform and if they help a lot with the motion transfer through the point point-to-form sometimes, because the point-to-form is basically a primitive wrangle that just takes primitive attributes and transfers them to the points and if you work with x, y, z, distance and primitive, you know what I'm talking about and if you didn't, it's alright, it's just a method of sampling data from primitives which is very precise and works really well with motion. So I'll be changing these parameters, I really want the mesh to be precise so I'm going to change the radius to 0 0 0 5 and the minimum points to 10 it's for more precision so if you template this yeah there's a lot of precision there 5 something like between 5 and let's say 50.5 changes if you do it like this the averaging of the motion is going to be less precise so I'm going to use a smaller radius and the minimum of 5 points this is enough for the motion transfer and another thing I want to do is transfer attributes. So I want to transfer the color because we have the color and the V. There's a bunch of attributes that Velom outputs but I want to transfer only the velocity and the color. So points to points, color velocity. So here we go. I'm going to sample this. So as you can see, there's color. What we're going to do with the color is apply the color only. This color will be used in the shader. So let's increase the sum account here. And the distance threshold, we're going to lower it down. So yeah. And as you can see, we have the velocity transferring to the quad mesh. Because initially, the pointy form doesn't do that. It doesn't transfer any attributes. So you have to do it by yourself. So the color attribute is going to be used in shading for blending in between two shaders. And... yeah. So... The normals... And what we're going to do else is an attribute cast. Like an attribute that I love a lot. It's just great. What it does is... It just takes any attributes and removes some digits from the data. data. So if you look closely in the spreadsheet here, let's see what it does. We have normal velocity and yeah, so you see the velocity has six digits after the dot here and the attribute cast and we'll type in VNCD here because this is what I want to trim some data. And I'm going to set it to 16 bit float. You'll see the difference instantly. So if you mouse will click on the attribute transfer, you see that the memory is 2427 and after the attribute cast it's 2268. This is what happened, it just trimmed a few megabytes which is great but with huge geometry and particles you can see a big difference in data amount, it gets trimmed. Nothing visually changes because if you disable or enable it you don't see any difference but if you look here we have six digits after the dot but if you do the spreadsheet on the attribute cast now, basically it lowered the precision. If we go here, let me find some intermediate. Yes, we trim some values here, add some other values there. but yeah it saves data on trimming values that are not really like it lowest precision basically so yeah we're gonna save this as a file cache so I'm gonna do it as explicit I wanna deal with the versioning now because I have too many versions in my life I've seen main clean pods, v0.1, F7, F5, VGLC. So I'm going to load it from disk. And we're going to simulate again. I'll see what I can do about the wiggling tail and come back to you about it. And explain to you why it does it. Resimulate and save this file cache. So now you have a recap of what happened there. I'm going to make this box smaller. So the vellum setup is split by groups here. We are a little bit over the organizing. So I'm going to separate this like this, pull it back, make a box for cleanup. This is going to be the cleanup. Cleanup and pits. Okay and now we're going to isolate the vellum here. The vellum is going to be in this area let's color it green film to set up now this is going to be the cleanup tool cleanup and motion transfer all right so we have some network boxes here And this one is going to be......couping. It's going to be......import and groups. Just so we keep everything organized, especially for people that are coming after you. Something pale, kind of like this. So now you have natural boxes everywhere. This one's gonna be on the side here. Sometimes it's really hard to follow through the tutorial. I know how it is. Sometimes I find tutorials that are not really explaining anything just clicking left and right, but there's no explanation why it's done and how data circulates especially in Hoody. You need that kind of explanation. There's a lot of stuff to consider and there's a lot of stuff happening under the hood. Once you get used to the data flow and how attributes circulate through those pipes, you start understanding stuff and debugging faster any problems that arise during simulation. Alright, so this is alright. We can save this. It's time to save. It's funny that I didn't save at least one time during this. This was really dangerous. So, yeah, we have to go and activate the autosave. So in the edit you click autosave and you'll have it autosave as many times as you set in the parameters and the options. So, yeah, next step. Once I save all of these files, the clean quad geometry and the resimulate this one. I mean, resimulate first and then save the clean geometry. Once I do that, we're going to switch to doing the FEM simulation with the wrinkles. the wrinkles and after this it's going to be the flip and after this we're gonna I'm gonna switch to shading our scene and rendering this bad boy. See you in the next one.",
+  "segments": [
+    {
+      "text": " So in the previous lesson I forgot to mention that I basically added a Vellum input output. It's this node. Vellum IO. I set it to explicit. I set my path, the frame range, applied the color to the geometry output, and emerged in the ground. So now we have this. And you can see what's the result of the simulation. looks pretty good for me. There's a little bug here, the tail that's wiggling a little, which is weird, but we're gonna fix it in the next iteration. But the compression of the muscles looks nice. The bubbles are not overly inflating, so I like it. We have this result because before I spent a lot of time tweaking the parameters and trying to dial in the values that work well, so this is why in the first long iteration we have this result. but usually it takes quite a lot of iteration just to test this simulation. So, and the parameters, we have the sliding of the fail here as we planned, then yeah, we have a bunch of stuff happening here. So yeah, then next step would be applying this animation to our quad mesh to see how it looks and use that quad mesh as a source for the flip simulation. This is not the whole simulation, it's just 115 frames, so it's enough for a preview. Sometimes I split the model into pieces, just for example I isolate one tentacle just to see if it works. So I'll give you an example. So basically if you have this as a model, sometimes I would come here, put a blast node in the middle, and I would select really a piece of the geometry instead of selecting this. So click again and select some points. Do this, did it not select it. And I would simulate only this piece of geometry. So yeah, this would allow me to have faster iterations. But it depends on the dynamics. Here we'll see just that the arm is going to be just throwing itself and then this part of the body is going to follow because we don't have any tail constraining the body to the ground. It's going to be a little bit chaotic. So it works in some scenarios, it depends. But yeah, I advise you to, when you try simulations like this, I advise you to go into your Microsoftres, disable a bunch of them and then simulate again. So I prefer to just simulate separately each Microsoftre, with each Microsoftre, like separate the simulation into several steps, see if separate things work properly. Yeah, so I'll disable this. Just to recap what we did here, we had a quad mesh. We grouped a bunch of stuff together. We painted some masks. We separated bubbles. We had a remeshed object. Then we animated it with the help of some attribute triangles. We transferred the animated attributes to our tetrahedral mesh. We had a two-out triangulated. Oh yeah, we copied some attributes, animated attributes to our triangular mesh and then, then shifted our triangular mesh with tetrahedralized it, transformed it into volumetric tetrahedrons. We transferred some groups, created an IDS attribute. We also set some casting properties. The vector called their arm, which defines the direction of the contraction of our muscles. We applied some density, stretch stiffness using our relative P-Box expression in the Z-direction, which allows us to have some really good control over our density. So if you stretch stiffness, and then we have density. So these are parameters that define how heavy the mesh and the object is in separate parts of the body. Then we transferred the live attributes onto our tetrahedral mesh. We attribute blurred. so it blends better between the vertices. Then we had a master transform that defines the initial placement of our monster above the ground. So you see it's barely above the ground for the initial drop. Then we set some valent cloth constraints which define the bendiness and the stiffness of the surface triangles. Surface triangles, if you remember, were created by the teticon form of this checkbox. So it's a separate additional layer on top of the tetrahedral mesh, which allow you to make some kind of wrinkles there, but if your mesh is not dense enough, you'll barely see those wrinkles. I added it just so I could show you the default, like the initial, the meta setup for these kinds of simulation. Then we created the Valentets. We applied some tetrahedral stretch constraints with a model called nonlinear ARAP, which is as rigid as possible, which you'll choice to maintain the volume tetrahedrons as much as possible during deformation and we applied those constraints to our group which is called GRP dense tetrahedrons only not the surface. The density was multiplied by a attribute which we created earlier, the density, then we had the thickness calculated uniformly over all of all the vertices. And then same with the stretch stiffness, we multiplied it by the stretch stiffness attribute. Then for the fibers we applied that fiber constraints only on the group called grp legs which is a point attribute, a point group, and the stiffness was set to zero because it's going to be controlled inside the valum-solver. So inside the valum-solver we created a sub-solver inside of which we're importing our rest geometry which is this geometry, the alt-rest which is geometry that has the live attributes cycling through. So in the sub solver we're basically transferring the live attributes onto the dock network geometry. Then we're sampling data from the ground. So before initially we as far as you remember we tried this locomotion test here in the subs just to see if the forces are transferred. We created a FRC debug force here which is the resulting force without any multiplication. We visualized it to see if it works properly, as we intended it to work. And then once we were done with the SOP test, we transferred the same VOP inside the SOP solver, right here, and we'll lock a motion test. So it does exactly the same, but over time. So this is the magic node, which just samples the velocity and the normal from the surface. The velocity is just a directional minus one vector in the z-axis. We normalize the forces, we added them together, multiplied them by CDR. The CDR controls the normal force if it's negative or positive over time. So the CD actually that activates the force only on the tips. And yeah, so the CDR is really crucial here. And then we safely added the force to the existing force. So the force here, we take the force here and then we add our resulting force to the existing force and we apply it to the force, not velocity. This is crucial because this is the safe method of adding. We can do it on the end velocity, but there's a lot of other stuff to consider. I'm not going to cover this here, but for now it's just, know that you have to, in this case you have to add it to the force. So yeah, so this is the subsolver. The subsolver pulls the live attributes and so they exist in the simulation, updated live on each frame. Then we created this geometry wrangle that reads the constrained geometry. So, the constraint geometry checks if the constraint is a specific type, which is a wrap, the as-readed as possible constraint. If it's not, it doesn't do anything. If it is, it just takes our derarm attribute and rotates it with the quaternion attribute, which is a default attribute contained in any tetrahedral mesh with the deadstretch constraints applied to it. The rest vector is the initial alignment orientation of the tetrahedron. So with the help of this rest vector, which is updated live during simulation, we rotate our V-dir arm so it points always into the direction of the deformation of our legs or arms or tentacles. However, you may name those. Then we have this, oh well, that would have been a mistake. If I scroll like this, it's going to increase the number. So Let's be careful there. We created the Vellon constraints Microsoftr, which is creating constraints between the tips and the ground at a specific time. And then another Vellon constraint properties, which takes those constraints, processes them, only them, and checks if the color of the vertices attached to those constraints is bigger than zero. If it is, it's going to kill them off with the use of this dummy, remove property. And then we have the animate contract which takes the fiber constraints we created earlier in the sub context. And it takes the color of those vertices related to the fiber constraints and applies a stiffness to them. So if it's white, it contracts, if it's black, it just releases. It doesn't expand, it just releases. So it removes the contraction basically. So the muscle becomes relaxed. Then we created another Bellon constraint attachment, attached to geometry. After the frame 5, between the group tail ends, which is the tail of our creature, and the ground force, we set also the crucial thing here, the sliding rate, which allows those constraints to slide along the surface. They're attached, but they have some freedom. If the sliding rate was zero, they wouldn't have the possibility to move. So the yarns will pull, but the whole body would stay in place because these constraints do not allow the creature to move. And finally, the bubbles animate, which is also a well-constrained property. Microsoft, which takes all of the text-stretched constraints, which is the whole body, and then it checks if those constraints are related to vertices in the group called group bubbles. If they are, then there's a fitting happening, which is taking the color of those bubbles and fitting them in between one and the max inflate value, which is 3, so between 1 and 3. So when the bubbles are black, the inflation is 1, which is basically the default state, the uninflated bubbles, the rest bubbles, and then max inflate is 3. So the rest length of these triangles and the tetrahedrons inside is going to multiply by 3, relatively speaking, because there's connectivity involved, involved so sometimes they do not inflate to 3 because there's limits where those constraints try to stay as really as possible so they do not allow inflation above some values. If we were to use another type of constraint like for example, if we go to this to linear A-Rap and uncheck preserve volume they would be able to expand almost limitlessly but the stability of the simulation would be completely different and there might be things that would explode. So here we go, we have this setup. I'm pretty happy with the result. This setup is quite clean compared to the initial setup I had before was just a horror because it was kind of a playground for me. I was playing around with nodes but this is a really condensed and compressed and a really short version of the tutorial. I mean, not short version of the tutorial, but it's the essence of the locomotion system I developed. So yeah, let's see what we do next. I will just apply, in this tutorial, I'll just apply the motion to our quad mesh and then we'll see what we can do next. So let's pull our quad mesh. Going back here to the top, creating an object merge, this into the object. This is how I like to do the merging. And yeah, so into this object, now we have this quad mesh, but as you can see it's penetrating the ground. And this is because we have master transform to respect. So you have to respect this master transform. So we'll just have to apply the same master transform to the quad mesh before we do any transformation or motion transfer. So when you right click on the master transform, and in the action you do a create reference copy. So this is a trick where you can transfer the same positioning rotation as our tetrahedral mesh. And now we're going to drop a point deform. And the first input goes our tree mesh. And we're going to drop a time shift. The time shift, delete the channels. So it goes into second input, and the animated version of the tetrahedral mesh goes into the point deform. So yeah. And as you can see, we have our quad mesh moving. And it's cool. One thing to mention is that the output mesh, as you can see, it's tetrahedrons. So what I'm going to do here, just clean the mesh. We're the only thing I'll be leaving here, is remove all the attributes except V and CD. and CD. So I'll uncheck the color because it was for previous purposes, not for the final result. And we may just, yeah, this, we have our mesh here. And then let's see what I'm going to do here. just bake this mesh so we can have faster feedback and yeah we have tetrahedrons here so one thing to mention is you see that we have polygons so we have tetrahedrons what I prefer to do is always do the point deform from the tetrahedrons basically the motion I love when there's tetrahedrons transferring motion to a poly mesh because the tetrahedral mesh contains volume volumetric tetrahedrons and they deform and if they help a lot with the motion transfer through the point point-to-form sometimes, because the point-to-form is basically a primitive wrangle that just takes primitive attributes and transfers them to the points and if you work with x, y, z, distance and primitive, you know what I'm talking about and if you didn't, it's alright, it's just a method of sampling data from primitives which is very precise and works really well with motion. So I'll be changing these parameters, I really want the mesh to be precise so I'm going to change the radius to 0 0 0 5 and the minimum points to 10 it's for more precision so if you template this yeah there's a lot of precision there 5 something like between 5 and let's say 50.5 changes if you do it like this the averaging of the motion is going to be less precise so I'm going to use a smaller radius and the minimum of 5 points this is enough for the motion transfer and another thing I want to do is transfer attributes. So I want to transfer the color because we have the color and the V. There's a bunch of attributes that Velom outputs but I want to transfer only the velocity and the color. So points to points, color velocity. So here we go. I'm going to sample this. So as you can see, there's color. What we're going to do with the color is apply the color only. This color will be used in the shader. So let's increase the sum account here. And the distance threshold, we're going to lower it down. So yeah. And as you can see, we have the velocity transferring to the quad mesh. Because initially, the pointy form doesn't do that. It doesn't transfer any attributes. So you have to do it by yourself. So the color attribute is going to be used in shading for blending in between two shaders. And... yeah. So... The normals... And what we're going to do else is an attribute cast. Like an attribute that I love a lot. It's just great. What it does is... It just takes any attributes and removes some digits from the data. data. So if you look closely in the spreadsheet here, let's see what it does. We have normal velocity and yeah, so you see the velocity has six digits after the dot here and the attribute cast and we'll type in VNCD here because this is what I want to trim some data. And I'm going to set it to 16 bit float. You'll see the difference instantly. So if you mouse will click on the attribute transfer, you see that the memory is 2427 and after the attribute cast it's 2268. This is what happened, it just trimmed a few megabytes which is great but with huge geometry and particles you can see a big difference in data amount, it gets trimmed. Nothing visually changes because if you disable or enable it you don't see any difference but if you look here we have six digits after the dot but if you do the spreadsheet on the attribute cast now, basically it lowered the precision. If we go here, let me find some intermediate. Yes, we trim some values here, add some other values there. but yeah it saves data on trimming values that are not really like it lowest precision basically so yeah we're gonna save this as a file cache so I'm gonna do it as explicit I wanna deal with the versioning now because I have too many versions in my life I've seen main clean pods, v0.1, F7, F5, VGLC. So I'm going to load it from disk. And we're going to simulate again. I'll see what I can do about the wiggling tail and come back to you about it. And explain to you why it does it. Resimulate and save this file cache. So now you have a recap of what happened there. I'm going to make this box smaller. So the vellum setup is split by groups here. We are a little bit over the organizing. So I'm going to separate this like this, pull it back, make a box for cleanup. This is going to be the cleanup. Cleanup and pits. Okay and now we're going to isolate the vellum here. The vellum is going to be in this area let's color it green film to set up now this is going to be the cleanup tool cleanup and motion transfer all right so we have some network boxes here And this one is going to be......couping. It's going to be......import and groups. Just so we keep everything organized, especially for people that are coming after you. Something pale, kind of like this. So now you have natural boxes everywhere. This one's gonna be on the side here. Sometimes it's really hard to follow through the tutorial. I know how it is. Sometimes I find tutorials that are not really explaining anything just clicking left and right, but there's no explanation why it's done and how data circulates especially in Hoody. You need that kind of explanation. There's a lot of stuff to consider and there's a lot of stuff happening under the hood. Once you get used to the data flow and how attributes circulate through those pipes, you start understanding stuff and debugging faster any problems that arise during simulation. Alright, so this is alright. We can save this. It's time to save. It's funny that I didn't save at least one time during this. This was really dangerous. So, yeah, we have to go and activate the autosave. So in the edit you click autosave and you'll have it autosave as many times as you set in the parameters and the options. So, yeah, next step. Once I save all of these files, the clean quad geometry and the resimulate this one. I mean, resimulate first and then save the clean geometry. Once I do that, we're going to switch to doing the FEM simulation with the wrinkles. the wrinkles and after this it's going to be the flip and after this we're gonna I'm gonna switch to shading our scene and rendering this bad boy. See you in the next one."
+    }
+  ]
+}