The Bordello: Calarel - Retrospective (Patreon)

It's been a while since I've done a retrospective, but I figured this would be a great animation for one. For this I'm just going to go section by section and hopefully provide some insight on the goals and challenges with each one, along with whatever else might cross my mind.

This is a very long post. So I've put together a little glossary for you. Unfortunately I have no page numbers to list but it should give you at least an idea of the sections below:

1. Visuals

1a. Transparency

1b. Volumetrics

1c. Successes

1d. Visual Conclusions

2. Physics Simulations

2a. Cloth

2b. Soft Body

2c. Jewelry

2d. Simulation Conclusions

3. Writing

4. Recording

5. Intro

5a. Scene 1

5b. Scene 2

6. Sex Sequence

7. Outro

8. Conclusions

Visuals

Before I go through anything else, I thought I'd do a section on one of my main technical challenges for this project: working with Blender's Cycles ray tracing rendering engine. For those that don't know, these days I mostly use Eevee - Blender's raster rendering engine.

A raster rendering engine is sort of like what you get in most video games. The lighting is essentially fake or approximated in order to get a much faster result. Ray tracing engines however 'simulate' the lighting realistically, by projecting light rays out and calculating their bounces and interactions. Much slower.

However over the years, the gap has closed significantly. Now-a-days I can make something in Eevee that people are convinced was rendered in Cycles. Even video games are incorporating very efficient forms of ray tracing in combination with their raster techniques. Indeed, even Eevee is incorporating such things. This is all great, but there are some things a ray tracing engine can do far better. These things include sub-surface scattering (or 'SSS', light scattering within a surface such as skin), refraction (light redirecting through a surface like glass), translucency (similar to SSS), volumetrics (fog or density in a liquid), complex bounce lighting (lights that bounce off a surface more than once), and motion blur.

A lot of the lighting for this project would have been much more difficult without a proper ray tracing engine, since much of it incorporated direct sunlight to then bounce and fill out a space. Plus there was tons of SSS and translucency with the plants and refraction with the glass potions and bottles. However there were some challenges that cropped up that I wasn't very prepared for that slowed me down significantly.

Transparency

Going into the project I had no idea that materials with partial transparency - such as the curtains around the bed - could create so much noise in the render. I figured that it would be slightly raised, but the amount of noise I ended up with was staggering. For those that don't know, noise is created in a ray tracing render when you haven't taken enough light samples to fill out a space completely. The more complex the path the light takes (generally), the more samples you're going to need to properly render it. More samples = more render time. And the worst thing about it is that even though it's not that bad/noticeable in stills, it can get way worse and way more distracting in motion.

Denoisers can help a lot, but even they can't figure out what's supposed to fill in a particular pixel if there's no information for it. So at best they smooth things over - losing out on detail, and at worst they turn the frame into a muddy mess as they try to figure out what the noise is supposed to be. They're great for smoothing over the little bit of noise you'll inevitably be left with from a render, but past a certain threshold they can only do so much.

Considering the curtains were comprised of two layers of lace fabric material with partial transparency, this sky-rocketed the amount of noise inside the bed. I tried a lot of tricks to cut down on it while not adding to the render times, but in the end it was only partially successful. I didn't want to re-imagine the set because I had already lit the majority of it and I liked the look. So the main solution was just to eat the larger render times. Ouch.

Volumetrics

This was a problem that I had seen coming, yet still failed to find a great solution to until late in the project. Much like transparency or refraction, volumetrics can generate a ton of noise and thus massively increase your render times. Though it's relatively simple and efficient in Eevee, it can be a major pain in Cycles. I had a few ideas for how to mitigate this going into the project, but with very little benefit. The main solution I ended up going with was just to use it in special cases, but for things like haze and fog I would just use Blender's built-in mist pass (which is more or less just a measure of depth within the frame) and some post processing. Near the end I learned a lot more about masking and compositing which could have given me a decent solution, but it was too late at that point.

As you can see above, the true volumetric solution would have looked a bit better by incorporating indirect lighting, properly filling out spaces, and being affected by direct light. However, it would have doubled or tripled the render times in order to smooth out a lot of the noise. Such long renders aren't a problem in small doses, but when you're trying to do a long project where you could be using this in hundreds of shots and iterating with re-renders, it can get out of hand fast.

Successes

If there's one thing I've learned it's that Cycles makes skin look really good. The difference in fake Eevee SSS and realistic random-walk Cycles SSS is subtle, yet huge. If you commented on how the skin looked nice, it's in part because of the renderer. It can be tough to dial in the SSS correctly, along with some other material values, but if you can get it in the right place then it works well.

I had first noticed this when I went back to cycles with my little animation involving Shadowheart and Lae'zel bathing. Before then, I just didn't really have the eye for the difference in quality. I could take or leave a lot of the other features, but proper SSS for close-ups is one of my favorite benefits for Cycles.

I was also able to pull off some shots that would have been very complex for Eevee, such as this shot above with Rose's potions. Each one is refracting light realistically, and with a little bit of cheating I was able to get that center potion to refract in an interesting way that worked well for the composition of the shot.

Visual Conclusions

So knowing all of this: was it worth it?

Cycles renders are going to typically take 3x - 10x (minimum) as long as an Eevee render, which can ripple out in a number of ways. Not only does it make the final render take longer, but it can make the iterative process take way longer too. There might be some stuff you didn't catch on your initial pass, or a lighting trick you wanted to try out, or a bug you didn't notice. Even just setting up the initial lighting can take longer because you have to sit there and wait for the details to display, instead of seeing everything instantly with Eevee.

It's also much less controlled. Light bounces everywhere, parts of the frame can get really noisy, some things can come through in the final render that you didn't see in the preview, etc. With Eevee - besides a few exceptions - what you put together in the viewport is more or less what you're going to get in the final render.

So for the future I can't imagine I'd like to use it on smaller projects, but for big ones it might just barely be worth it - especially since I'm much more familiar with it now. However it would help to have even more and even better rendering hardware than what I have right now, and I already have pretty high-end computers.

Anyways, now I can get into one of the other major aspects: physics simulations.

Physics Simulations

Physics simulations comprise a lot of the challenges and roadblocks for many kinds of animations, but they can also be the secret sauce. They can be so complex and vital that some artists specialize exclusively in them. For this animation, physics played a vital role and came with a variety of successes and failures. The three main aspects here were cloth simulation, soft body simulation, and jewelry simulation.

Cloth

Though it didn't factor into the sex scenes themselves, there was a lot of cloth simulation in the intro scenes. Darcey's dress, Rose's dress, and Calarel's cloak all necessitated some level of simulation and just couldn't be totally managed with hand animation. Done right, this can look great. However the challenge is in being able to do it in the first place. You need the right topology (not only for the simulated mesh but for the collider meshes as well), the right settings and vertex groups dialed in, the right animation to accommodate the simulation, and a whole lot of patience.

In preparation for this animation Axel and I put together a lot of clothing items. We only featured a few of them in here though, and we're still working on more for future animations. Each item had to either be built from scratch or retopologized off of a reference mesh, then painted, rigged, and calibrated for simulation. It takes longer to build up assets this way, but it allows you to properly prepare the assets for simulation. Certain kinds of topology can make it way easier or way harder to simulate, so everything was put together with this in mind. Calarel's whole outfit, Darcey's dress, and all of the extra's undergarments were original assets from us. The only thing not totally original was the footwear and Rose's robe, though I built her cords, jewelry, and headband from scratch.

So that part went smooth, but I ran into a few hiccups on the animation side of things. There were a few parts in the intro where fabric would get caught between legs or arms and start breaking. This necessitated a lot of trial and error and in some cases total reanimation just to accommodate the physics. Even past that, it was hard to dial in settings in Blender's native cloth simulator that produced realistic results consistently from shot to shot. They required frequent changing, which required even more trial and error.

I've been trying to avoid this, but in the future it might be best to just use something like Marvelous Designer for this kind of work. Blender's cloth simulations just have very clear limitations when it comes to collisions and it's possible another program could handle them better. It's a shame because it can add a lot of time and clutter to a project when you have to ship things around like that for simulation, but the alternative could actually take longer in cases like this.

Regardless, Blender's native physics were able to just barely hold up for this project.

Soft Body

Rose was conceived in part so that I could use her as a platform for a lot of physics simulations at the same time. Her breasts, belly, butt, balls, and thighs could all be simulated - not to mention her hair and jewelry. What I put together technically worked, but there were many occasions where it wasn't necessarily better than a hand animated solution.

It was hard to dial in her thighs and butt to give an appealing response in most poses. They're much more firm than the breasts and belly, so the range of movement you expect from them has to be much more precise. This firmness also makes them easier to animate by hand, since you're expecting less changes to their form overall. I was never really happy with what I could get out of blender's native simulator for these occasions, so there wasn't a ton of it that got into the final product.

The breasts and belly however worked pretty well. They both had collision with each other and required little dialing in. Since they're a lot less firm and more fatty, the range of movements we expect from them are much larger and more forgiving.

Calarel was a similar situation. She's thinner overall so I was only really concerned about simulating her breasts, and I had multiple options for doing so (cage simulation, line target simulation, wiggle bones, hand animation). Even her hair was fairly simple, with bones being used for the ponytail and cloth simulation for the fringe.

Overall I felt like I was getting the most out of what Blender could do, but was left desiring more in some circumstances. I'd like to look into more complex simulators for soft body in the future. Hopefully with a fairly simple workflow.

Jewelry

This one really kicked my ass. Going into the project I was expecting much of it to be pretty simple, but it turned out to be another animation headache. Rose's necklaces required a lot of trial and error until I settled with the easiest option of just simulating a card on top of her body, attaching a medallion to the card, and then animating on top of that. Not a great solution, but far better than the others I tried. It turns out simulating rigid and cloth physics together isn't so simple.

Her bracelets were by far the hardest. Apparently Blender has a lot of trouble simulating rigid objects that encompass their colliders, so my solutions going into the animation had to be heavily modified. In the mean time I decided to just animate them by hand, noting that I could at least get the results I wanted on the first try, as I was in no mood for more trial and error. It wasn't until I was near the end of the animation that I had discovered more robust simulation solutions for the bracelets. Next time, perhaps.

After all of the other stuff I had little patience for her earrings and headband. Those would have to be animated by hand or not at all. Not a great showing for this category on my part, and definitely a big time sink.

Simulation Conclusions

Though I am getting better at working with Blender's tools, it's becoming increasingly obvious that there are better solutions elsewhere. Unfortunately that means more tools to learn, more time spent exporting/importing, and more space spent on big simulation files.

I didn't really cover the fluid simulations in this section because for the most part those were done by Morfium. I think he did a pretty stellar job with them this time and I'm happy to let him worry about the nuances of x-particles. The stuff that he didn't do - mainly the first creampie - was just stuff that I hand animated myself.

Anyways, onto the story itself.

Writing

I don't usually talk about writing since I mostly try to give a cold, technical breakdown, but I figured I could have some potentially useful insight this time. Here I'll just briefly go over what my typical process is from writing to recording dialogue and how it went down with this project in particular.

Typically the start of any longer animation like this is a script. I can wing more action-heavy animations without much dialogue, but as soon as you incorporate anything more than a simple story with simple characters, a script becomes an important thing to reference - even if you don't stick to it.

The rule of thumb for screenwriting is that one page generally equals one minute of screen time. This can change according to what you're actually writing into the script, but it's a surprisingly reliable estimate. For this project the script was about 3.5 pages, with just the intro and outro. The intro was about 3 pages and ended up being 2:50, and the outro was a half a page and came in at 0:45. So yeah, pretty reliable.

I typically don't write anything detailed about the sex scenes just so I can give myself enough wiggle room to figure things out in the moment. Lately I've taken to writing extra material for it later once I have a better idea of what I can do with the models and what might work better. In this instance a lot of the sex dialogue was written and recorded later, or simply improvised.

When writing for something like this you have to walk a very careful balance. I find it much harder than writing longer scripts. At every turn you're facing down tough decisions. The longer your story scenes are, the higher the audience expectation could be for the sex scenes. If the proportion is too lop-sided, it could be confusing or disappointing. However if you cut things down too much, it could feel dumbed down or rushed. Want to have a nice, complex story with satisfying action? Be prepared for way more work, which means more time, which means even higher expectations.

Suffice to say: brevity is the soul of whit. These kinds of projects really exercise your ability to compress a complex narrative down to a dense runtime. Which is, coincidentally, good filmmaking practice. You really can't leave much fat or filler in these narratives, you have to be efficient, and you have to be using all avenues of communication at the same time: visual, auditory, and contextual.

So even though it might be nice to get an establishing shot of Rose's bordello, or a shot of Darcey and Calarel knocking on the door, or a shot of one of Rose's girls grumbling out of bed because of the noise, or a shot of Carlarel changing - they just aren't very efficient. Every line has to be well considered when writing, because those lines turn into hours and days of work during production.

For this animation I really wanted to at least tell a full story for Calarel, as opposed to just trying to characterize her. So I used as much time as I could setting up and paying off her journey. From that hesitation starting with just a little prick, all the way to her realizing she might have made the wrong choice just a little too late.

Recording

After multiple script drafts I arrived at something I liked, and contacted the voice actors I wanted for the animation. Two of them were returning - Kassioppia and Baku Satsu - and one was new but someone who I've worked with before - Bordeaux Black. Having worked with everyone before, it was a pretty simple process. Baku fell back into the character pretty easily, I developed Calarel for Kass a little more, and I put together information on how I wanted Rose to be for Bordeaux.

For me, this is one of the most enjoyable parts of the whole process. Working with actors, fleshing out characters, and honing in their designs is great because there are no hard edges yet. Unlike 3D, you get to be creative without immediately feeling the consequences.

From there it's just about going back and forth, providing direction, tweaking the script, and improving performances until you're happy. This process can be pretty slow as it's often done remotely over the period of days and weeks.

After all that I retreat back into my hole and get back to work, occasionally popping out for some more recordings. Then at the end when the animation is nearly finished, you have all the footage you need for the performers to dub to. Dubbing is just the moans and groans, along with any improvised dialogue that could still make it into the product.

None of this takes up nearly the amount of time it takes to actually animate the damn thing, so you treasure it while you can. Plus you can more or less do it concurrently with other tasks, so proper planning can mean that you're able to give your performers lax deadlines and more room to give a proper performance as you'll have other stuff to work on in the meantime.

Speaking of, now it's time to get into that 'other stuff'.

Intro

Scene 1

Generally animation that isn't sex or action related is a lot easier and takes less time, but this scene wasn't quite so simple. For one, this scene was going to be my main challenger for environmental design, due to the complexity of the environment and the lighting challenges that came with it. Two, there was a lot of unorthodox animation, like turning and walking, walking on uneven terrain, stepping forward and stopping, etc. Things are easier in cycles - or loops - when it comes to animation, so anything that breaks up those cycles can add difficulty to an animation. Even though this still went way faster than the sex scenes, it did cause me to stumble at parts.

Initially when planning the environment, I wanted a self-contained area that was striking yet still foreboding. Even during the script writing phase I had some idea for how I was going to block it out - with Darcey and Calarel being in the middle of the courtyard and Rose looking down on them from above. All of the straight vertical lines surrounding the manicured garden were meant to give this kind of restrictive, jail-block vibe without making it seem like an actual uncomfortable place to live.

I spent a lot of time on the environment - probably too much - so there was a fair bit that didn't get shown, but at the same time it made for good practice. I think I accomplished my design goals for it, but if I were to try again today then I think I could get it looking a lot better. The man-made assets could be better nested in the environment, the topology on a lot of the assets could be better, the foliage could be more natural, and the set as a whole could be a lot more detailed and less 'sterile'. I think next time a much more heavy use of reference could help a lot, and designing the set around shots as opposed to making one big unit could save me time and give me a better end product. All that being said, I was at least quite happy with the shot of the rose.

Now for the animation. I had a fairly easy time blocking, laying out all of the keys, and lighting each shot, but I had run into a little roadblock. I had initially planned to have another animator do all the in-betweens and the polish, but things fell through and it came back to me. Already saddled with a lot of work, I had to put it off and it just ended up making the project take longer. I hadn't done a great job planning some movements out and it really made life a lot more difficult for me. Some stuff like the conversation between Darcey and Calarel on the bench was a breeze, but other stuff like Calarel sitting down or Darcey grabbing her hand posed both animation issues and novel issues.

Characters interacting and characters moving is one thing, but characters interacting while moving around is a whole other thing. Without getting too much into the details, I had failed to come up with an easy solution to resolve conflicts between both of their pose and object spaces from the blocking phase, which just meant I had to spend more time making everything line up. Combine this with doing some genuinely unique movements that I haven't done before, and it became a challenge. Pretty minor, but still a challenge.

Another headache was the cloth simulation. Darcey was fairly simple with just her skirt being animated, but Calarel's cloak proved to be a real pain. In most normal cases it was fine, but if it got stuck between the arms it turned into a problem. I'm still not too sure how I want to go about fixing these edge cases in the future, but for this one I had to re-animate some sections just so the cloak didn't get caught in anything. Which required tons of cloth simulation iteration. In the future, more and better colliders combined with some force fields might be more robust solutions.

The last problem I was discovering more or less throughout the whole animation was the use of sun lamps. In Blender, they're essentially tiny pin lights that simply have to be rotated and project an even amount of light no matter what their position is. Essentially: the sun.

It turns out, the sun is not a tiny pin light. In fact, it has an angular diameter of roughly 0.5°. At the time I had not realized any of this, and unwittingly had the angle set too low, or had neglected to turn it up higher when it would have benefited me stylistically. This resulted in the lighting in many of my shots being a little too sharp, which generated some bad shadows and revealed some bad geometry. Oh well, this is why I need more outdoor scenes to help me figure this stuff out.

Grumblings aside, I still like how some things went. The close-up of the hands at the end of the scene, Darcey's gesticulations, and the general environmental animation was pleasant.

Scene 2

Scene 2 was for the most part pretty nice. I only really cut two lines of dialogue that I found redundant, and Rose's de-robing was somewhat abridged, but the rest went down about how I envisioned. In this scene I was most worried about the potion shot in the beginning, and Rose's pacing around Calarel in the middle. However, they went surprisingly smoothly.

What was neat about the potion shot as well as the mirror shot was having to grapple with real world light physics. Both shots required taking into account reflection and refraction when setting the focal plane, which had impacts on the focus pulls that occurred in each shot. This is just stuff you don't really think about unless you're working with a ray traced renderer.

 The pacing section was its own set of challenges. For one, animating a character that's pacing as well as interacting with a still object just comes with its own problems in terms of constraints. Two, I had animated this after the first sex scene, so I had to match the lighting. And three, I had to block, animate, and shoot this in a way that allowed Rose to do one full rotation around Calarel so that she could hit her marks at the beginning and end of the exchange, which would be a challenge even for an actress in real life.

Thematically, this section was the culmination of their introduction with each other. I wanted Rose to be sort of like a predator hunting her prey. At first she scouts her prey from afar - we don't get a great look at her. Then she gets closer and closer, and we get better glimpses of her yet never a full close-up. Then she circles her prey, prowling around her. Finally, the first time we get the close-up is when she finally pounces and pushes Calarel onto the bed. It might not be explicit, but it helps give the animation some perspective and gives Rose a bit of a menacing feeling.

The lighting turned out to be doable, though unconventional. I made tweaks that separated the background and the foreground a bit, but it wasn't too overbearing. I also liked the idea of cutting the light off half-way down Calarel's face, as it not only helped focus the audience's eyes elsewhere, but also yielded thematic results by covering Calarel's eyes in darkness. All of this also helped to prevent the frame from getting too blown out from the intense direct light.

Though the best surprise was that the animation turned out to be much easier than I thought. I just had to figure out the right set of constraints. Since we never got a shot of Rose's feet, I never had to perfectly position them on the ground, which allowed me to prioritize the timing and blocking of her movements. The only loose end was that her right hand would switch a constraint to her body on and off when she wasn't and was touching Calarel, which would cause its own issues. Luckily though, no huge set-backs.

Finally the lighting match between this scene and the next worked a lot better than I thought it would. So, no big regrets with this one.

The only thing that was somewhat cut short was Rose's disrobing. I had initially thought of it as being a big focus, but when laying it out that just didn't make sense. How was she going to take off her belts? Are you going to rely on the finger colliders to move the robe straps off of her shoulders? How do you know it wont break? Will it even fall off? Are you ready to simulate this 300 times? Bleh, the feet shot is fine.

Sex Sequence

Instead of just doing things the traditional way, I wanted to order things a little more unconventionally. The big thematic push here was to make it feel like Rose was really using Calarel, as opposed to letting Calarel prove herself as a performer. As a result, Calarel is rarely put into a position where she can be proactive, and is instead simply reacting for most of the sequence. The jerking scene was a great way to introduce the dynamic, even though it's typically how you'd end a sequence.

This one was also the hardest to light, since I wanted to go for something really unconventional. It's essentially an all-fill, no-key, no-back light set-up. These kinds of set-ups are very hard to pull off with a light source like the sun because it's very hard to direct the audience's eye and separate the subjects from the background. Thus, I had to utilize a lot of texture in the lighting - and a lot of flags.

Flags in filmmaking are used to essentially block light - they're usually just black rectangles. And boy did I use a ton of them on this animation. The benefit of using them in 3D however, is that you can make them invisible to the camera. For this scene I had used them to block out the light in the background, light on Rose's thigh, and diminish the light on her upper torso. So even though the impression is that they're all getting evenly coated in light, it's really focused around particular areas.

Now a lot of this has been made easier due to light linking, but I didn't have access to that at the time because Blender hadn't implemented it yet. Grumble.

The animation itself was pretty straight forward - nothing I haven't done before. I just needed to make sure I was creating a good animation and lighting environment to allow Morfium's simulation to really pop. That meant anticipating what it might look like and calibrating Calarel's reactions accordingly, and adding some supporting lights that help shine specular highlights onto the jizz so that it doesn't get blended away by the intense light.

The kissing/dry humping scene was a whole different can of worms. The lighting and animation wasn't a big deal, but the topology was. To really execute on this I needed to create an entirely different, higher-poly version of Calarel that allowed for more detailed butt deformations.

So first off, why wasn't she super high poly in the first place? Well much like game models, 3D animation models need to optimize too. Though they're higher poly than what you might find in a game, too many polygons can pose problems. They can slow down viewport performance, take more time to process with modifiers, and increase the disk space cost for things like shape keys. So typically there's an upper limit to what most animators are willing to accept.

Many will opt to have high poly and low poly models of the same character, so that they have something efficient and optimized to work with without sacrificing quality. For NSFW work like this however, character interaction and deformation is very important. You might get some good use out of a low poly model during the very beginning of the animation phase, but the bulk of your animation work is still going to be spent on the details of the character interactions. Hands grabbing thighs, hips hitting hips, lips on lips - you get the picture. Since you're spending so much time doing this stuff anyways, many people just forego the lower poly model entirely and try to build something more universal.

For this situation however, I needed many more polygons, which meant I had to figure out a way to increase Calarel's poly count without messing up... everything. It sounds easy on paper, but this is Blender we're talking about.

If you try to subdivide the mesh (multiply the polygons), you're going to run into several issues:

1. The shape is going to change. If the shape doesn't change, you have bigger problems.

2. The weight paints are going to be affected.

3. The color attributes are going to be affected.

4. The shape keys aren't going to transition properly.

5. Several modifiers, including particles and mesh deformations, will likely break.

In an ideal world when you need a higher poly mesh, you just duplicate what you have, click 'subdivide', and move on with your life. But for a situation like this, I needed some time and a special add-on to subdivide the model without running into every single problem (and more) I listed above.

After much trial and error, I finally found something close enough. The only problem was that #4 and #5 weren't entirely solved for, but it was something I could work off of. So the solution was to use this high poly model for those hand-to-butt interactions, then simple render that mesh separately and composite it onto the lower poly mesh that takes up the rest of the frame. All for a few slaps and grabs.

Add this to the pile of reasons I ended up retopologizing my models later.

The missionary pose was one of the easier scenes for me, though still quite time consuming. The main hitch here was just the vagina animation. There's a lot to account for, and even an automated solution would require some hand animated touches.

Going in, I didn't realize the vagina rig was in such disrepair. Instead of totally redoing it, I had the idea of just adding on top of it to increase the functionality. It was an... okay solution, but at a level of jank I just wouldn't recommend to anyone else. The end product looked pretty nice though, and that's what matters most.

The jizz in this scene was all hand animated and fairly easy to put together too. I've had a lot of experience doing that kind of work at this point so it's something that I can easily plan out and just crack down on.

Doggy was definitely the easiest. Very animation-intensive still, but a lot more fun since I didn't have to worry about many details. No weird shadows, not much penetration, not a lot of interaction, just plain and simple.

Since one of the downsides of cycles is longer render times, it did make it a little harder to hone in the jiggling and rippling. You usually never know quite how well its going to turn out until you see it lit and fully rendered with all of the modifiers applied. With Eevee it's a lot easier to just render, iterate, render, etc. But with cycles everything just moves much slower.

On the other hand, this was a nice opportunity to display all of Rose's physics, which came together pretty well after some tweaks.

Similar to the doggy scene, this one was pretty simple. The biggest challenge was just in posing. Usually you'll have something in your head before you start that's roughly close to what you finish with, and the posing process is just about trying to work out the real world practicalities.

I liked the lighting when I set it up, but after seeing it in motion I started to notice some key issues with the sun lamp itself and with the SSS of the characters. Another disadvantage of Cycles is that it makes spot-checks harder as well. You can't really play back the animation in the viewport with the renderer on and expect any kind of reasonable FPS. So you have to skim, and when you skim you're prone to missing weird shadows or mistakes. Even then, unless you do a full render you can miss some finer artifacts like the SSS being too intense in some areas. Since those areas tend to take longer to render, they don't fully reveal themselves in a preview.

Now this right here was probably the most difficult pose of the bunch. The only thing that was relatively simple about it was the lighting.

To start, the characters were all over each other and nearly constantly pushing and pulling on one another. This creates the need for a lot of constraints, restricting the freedom of movement you're allowed in a pose and creating a lot of potential clipping problems.

Second, there was penetration almost constantly showing and at very high speeds, meaning that things could very easily break and clip.

Third, there's never a very clean break between either character's hips, so the action and reaction is less defined and harder to accentuate.

Finally, Rose's hand on Calarel's butt required a higher poly count. However, the technique I used to subdivide her mesh before never really transferred the shapekeys and weight paints perfectly. So in a pose like this where she's twisted up like a pretzel, there are loads of driven corrective shapekeys making sure the mesh looks right. At rest, the subdivided model might look the same, but when things are this twisted up all of the minor differences in the weight paints and shapekeys really start showing.

So I decided to just stick with a multi-res sculpt instead and keep it simple. Then I would composite that on top just like I did with the kissing pose. Unfortunately that didn't work quite so easily because of a variety of reasons, including that jury-rigged vagina rig I set up for the missionary pose before. Things started catching up with me.

The nice part about short animations is that when you're done, you're done. You can make whatever stupid decisions you want and then never touch the project file again. In projects like these, you incur a debt.

1. There's a tech debt from crudely modifying and using the same models. Ex. You did a shoddy patch job for a bunch of glitchy vertices on your model instead of solving the root problem, and now they act up every now and then.

2. There's a continuity debt from including something in one scene that you then necessarily have to include in the next scene. Ex. A character gets scratched in the face, now she has to have a scratch on her face for the subsequent scenes.

3. Then there's an expectation debt from including something that the audience expects will develop into something else. Ex. Your project involves multiple sex scenes, that means it must include a money shot of some kind.

All of these debts are what can make longer projects exponentially more difficult, and why many people don't do them.

So long story short, there ended up being a lot more rendering then I'd like and a lot of brute force work on the compositing end. But I got it done nonetheless.

 Then there's this scrapped scene. I had set it up and animated it early on, but at the time I just didn't really have the technical expertise to get it done properly. I wanted to rely more on physics, but I really didn't have any of the techniques figured out yet. The shots would only last a few seconds before things would start to break. Realizing I had no viable solutions at the time, I decided to replace it with something else.

I think I could potentially get something like this done now, but it would take a lot of rigging and maybe some re-imagining.

In comes the mating press pose. These are usually pretty fun and not too challenging. Since I already had another pose that was similar but slow and powerful, I wanted to try to go for speed on this one. As a result I pushed things about as far as they could go. Looking back I think I could do a little better.

When you're doing really fast, repetitive motion like this, you have to take special account of the framing. When the action takes up more of the frame, it can get harder to find something to focus on and you end up getting lost. You also just get way less time to build up anticipation, communicate follow-though, and prevent the whole thing from feeling machine-like.

Adding more variance, breaking up the frequency, and reducing the range of motion for the close-up shots probably could have helped here for the first section.

Finally... whatever this position is. It's quite a complicated pose and a good way to cap off the sex sequence. The main challenges here were about making every angle look good at the same time and getting the creampie to work in a way that made sense.

Usually when you're putting together a scene, you have the option to treat it like it was a multi camera scene or a single camera scene. In film and TV, when things are shot multi-cam it means they have multiple cameras recording the same take simultaneously - think live studio audience shows like Fresh Prince of Bel Air. Single camera obviously only involves one camera - think more dramatic stuff like Inception or Breaking Bad. There are benefits to both, and those benefits and drawbacks can translate over to 3D as well.

When constructing a pose, you're never just throwing it together in a void without any other considerations. You're putting it together with context in mind. Where's the camera going to be? Where are the light sources? What range of motion will they need? How does this interact with the environment? Etc. If you want the best looking pose, you're typically going to pose around a singular camera angle so that you can bias towards it in the most ways possible. However, if you want more coverage, you have to leave room for other possible angles too. Bias too much towards one angle, and you might break your models too much to allow for other angles without a bunch of extra work. Don't bias enough, and your pose can come across as flat and uninteresting.

For this one, both the wide angle from the side and the close-up from behind had to be equally great, but compromises had to be made for both to make either work. All this to say: there was a lot of trial and error before I was able to find a happy medium between the two. I did have to cut some other angles in the process, but in the end it was best for the scene.

The creampie was another thing I had Morfium's help with. Though it was hard to figure out just what it should look like exactly. Realistically it would be rather unimpressive, most likely just a stream down the middle. That didn't really seem explosive enough for a finale.

You often come across these problems in porn. For things to be grounded, it's often best to be subtle with them or not show them at all. But to push the appeal of other things, you need to think about what it could look like. This is where a lot of tropes from porn come from - when an artist figures out just how to navigate that thin line between appeal and absurdity. Artists that do niches like me often joke about how much easier it would be to do regular vanilla material, since it doesn't require nearly as much careful navigation. As someone who's done vanilla material, I can confirm this is the case.

So after conceptualizing it, we worked on dialing it in until we got the right consistency and movement. Since most fluid simulation at the moment has problems with not leaving much residue, I tried to help it out a little bit by using dynamic paint to drive a jizz material - to limited success. Combine all that with a little compositing and some stellar voice acting and it turned out alright. Had we done it again I'd probably dial it back a bit but that's a lesson for next time.

Outro

After all that, the outro was the last thing I did and by far the simplest. I just had to figure out a rig for the scroll and how the whole stack of clothing was going to work. I also changed the lighting and color a little bit to communicate that we had moved from morning to afternoon, and that Calarel had been put through her paces so to speak. The animation was simple as it was all just close-ups.

Conclusions

This right here is definitely the longest... anything... I've ever written, but hopefully its provided some insight for those who have actually managed to get this far.

This is also the largest animation I've ever worked on. Over a hundred lighting set-ups, thousands of renders, and millions of keyframes.

Animations like this are anything but easy, especially when you're almost entirely doing it on your own. Every mistake, every hitch, every glitch, every challenge and every decision ripples out through every day, week, and month you have to work on the project. If you're not stubborn or motivated enough, it has a way of swallowing you whole. Which is why pacing, hobbies, and side projects are so important.

At the end of a big project, every artist will probably tell you the same thing: that they don't want to look at it anymore. It's not because they don't like it, but because they're just done with it. And I am happy to say that I'm done with this project. I've already internalized all of the lessons and carried forward what I wanted for the future. If you were to ask me what I thought about the animation itself, I'd say: "Pretty good, I guess."

And now that I'm about a week out from the release I am yet again, as naively as ever, convinced that the next project will be easier and better than the last.

Thanks for reading!