nyl2 - Girl's Night Out - Retrospective - Free Hentai Manga, Doujinshi and Comics Online Reader | LewdVault.com

It's that time of year again - the time where I get to look back and ask what the hell happened.

I don't expect this will be quite as long as my Bordello: Calarel retrospective, but there's still quite a bit to cover. Since the animation is a bit unconventional in terms of pacing and intercutting, I'll be taking a broader approach and therefore might cover things out of order.

Here are the sections I'll generally be covering:

1. Visuals

1a. The Club

1b. The Bathroom

1c. Mirrors

1d. Lights & Reflections

1e. The Hallway

1f. The Security Room

1g. Graphics

1h. Desktop

2. Animation

1a. Cloth Simulation

1b. Soft Body Simulation

1c. Fluid SImulation

1d. Poses

3. Editing

1a. Layout

1b. Sound

4. Conclusion

Visuals

This was quite a welcome change from the candle-lit fantasy settings I'm used to working with, so I really tried to lean into the technology as much as possible.

When developing the concept for this one I wanted to keep the found footage/handheld aesthetic of my Tracer and Widow animations and push it as far as I could. So I decided pretty early on to commit to having all 'practical' cameras. I'd use whatever I could (CCTV, phone cameras, webcams, hidden cameras, etc.) to capture what was going on. Over time this evolved into being incorporated more into the storytelling.

Given the opportunity and the filming techniques, I wanted to lean into stuff I wouldn't get the chance to do in any other project. So it was all light panels, spotlights, mirrors, sleek designs, giant screens, and so on.

The Club

First that brings us to the club itself. The interior was designed as a giant atrium with several tiered decks, VIP areas, a central dance pole, and - fun fact - even a dance floor that I never really got a shot of.

This was sort of an exercise in optimization. I could get away with sleek, simple designs because of the sci-fi nature, but it still had to be populated and alive. The thing was that I wanted to keep it very simple to run in the background, so it was all about designing repeatable, easily manipulated, low level of detail assets. The structure itself was easy enough, but the crowd, details, and spotlights required a bit more attention.

Details like railings, tables, lounges, etc. were just low detail versions of assets I had already made. And populating them was rather easy with point distribution in geometry nodes.

Spotlights were a bit more complicated and had to be proceduraly animated using sine-based drivers through the geo nodes themselves. They were emitters but didn't actually cast light. For the light cone I just used a literal cone - just something like 12 polygons. A transparency gradient made it look well enough like a volumetric without the cost or set-up.

Crowds were a whole thing. For distant crowds I used a slightly more sophisticated 'billboard' approach. I took a single image of a row of silhouettes, cut it up into different meshes, and populated them using geo nodes. From there I just animated them using similar methods to the spotlights. For such a simple approach, I think they turned out pretty well.

Obviously this solution wouldn't work up close, so I turned to Procedural Crows - an add-on. Essentially what they do is a more sophisticated method of what I was doing, but with built-in low-to-mid poly characters and some stock animations. It turned out to be a pretty great tool and I only had to make a few tweaks to make it not too glaring. First I did a blanket material replace on all of the models to a sort of black so they wouldn't distract you too much. Then I just tweaked some of the animations and poses so they didn't look too alike. This is one of those aspects I could have spent a hell of a lot more time on (using models with more overwatch-like proportions, making my own custom animations, making them more active/reactive), but I was content spending that time elsewhere.

Oh and the phone lights were just a bunch of emissive particles floating around with a depth-pass-based streak glare composite over top of them.

The Bathroom

This is one of those environments that is literally no sweat for a ray-traced engine, but an absolute nightmare for a raster engine. I covered this in my Bordello: Calarel breakdown, but I'll quickly recap here. Blender has two main render engines: Cycles and Eevee. Cycles is a ray-traced engine - which essentially calculates light realistically. The heavy-duty stuff. Eevee on the other hand is raster - like a game engine. It uses all sorts of tricks to calculate things in real-time.

I mainly used Eevee for this project, so the raster aspect provided a few key challenges in this section:

1. Raster engines do not play well with mirrors.

2. Raster engines do not play well with large, complex light sources outside of view.

3. Raster engines do not make it easy to evenly light areas.

Now Eevee has actually received quite a bit of improvements over the years that made these challenges easier. They've implemented things like partial ray-tracing and light probes to help account for things outside of screen space. However these aren't complete solutions.

Mirror reflections aren't 1:1, and often drop details and specular reflections. Emissions outside of the frame can still 'pop in' lighting when they enter frame. Light probes can have blind spots that cause weird reflections or shadows.

One option was to just lean on Cycles for these aspects, but I wanted to challenge myself to do it all in Eevee. So, how did I cope?

Mirrors

If I wanted to capture the reflections in the mirror in complete detail, I knew I'd have to composite one rendered frame on top of another. So I needed a unique rigging solution that essentially created a 'mirror camera' to capture the characters from the exact opposite side. Then I would simply overlay one over the other. It would require precision.

What I ended up with was a 3-point set-up. The first point - the camera point - would copy the rotation and location of the main camera. The second point - the pivot point - would act as the single, stationary point exactly between the other points. The third point - the mirror point - would translate and mirror the movements of the camera point by using a custom space from the pivot point. Then, the 'mirror camera' would be parented to the mirror point and inverted on the x-axis. The end result: a completely mirrored view.

Then I just had to make sure the assets in front of the mirror lined up perfectly and equidistantly to both camera views.

I also used this technique in the dressing room scenes. Here's a side-by-side of the composited final result and Eevee's probe-based reflection.

Lights & Reflections

So that was a the hard problem down. Now is was just about controlling the image in motion.

I had essentially three tools to help limit flashing, flickering, and lighting inconsistencies: Probes, artificial lights, and resolution.

1. Probes are Eevee's way of compensating for screen-space limitations. Even with their partial ray-tracing implementation, they still mainly stick to calculating what's on the screen at the time. Sphere and plane probes can help with reflections. By positioning them on or around objects, it indicates to the engine that you want them to reflect what's beyond view. Volume probes help calculate and bake light bounces that are coming from off-screen practical lights like emissives (in this example, the ceiling light panels).

2. Artificial lights (as opposed to practical lights) can be placed to mimic the light that you'd expect to get from an off-screen source. It isn't a perfect solution, but it's the oldest and most basic form of dealing with the problem.

3. When in doubt, zoom out. By proportionally increasing the resolution and the focal length of the lens, you could essentially increase the render boundaries of your frame. It would take longer to render, but it would be able to catch and fade in these elements before they entered the final, cropped frame.

With all of these tools combined I was able to (mostly) control the lighting in the frame so we could have a sleek, well-lit interior with tons of ceiling lights and few inconsistencies or pop-in.

The Hallway

This part was all about the art. My main challenges were the story-specific assets, the massive amount of graffiti, and the titles/logos.

I knew I needed these posters and stickers earlier on, so I hit up an artist that I thought could create pieces in a similar style to Overwatch's in-game posters. They had sort of a simplified, vector-art style to them. At the same time, they needed to be sexy. So I figured honeybread could do the trick.

I was pretty happy with the results and here they are if you want to check them out in detail.

The graffiti was a different challenge. I had looked around for graffiti covered walls, or transparent graffiti collages, but nothing was really up my alley. I needed a certain aesthetic and a certain level of control. Luckily I came across a very large library of actual scanned and processed graffiti. I just had to process it, bring it in, and populate it. Easier said than done, but once I got all of the vector and z-fighting problems out of the way I... maxed out my VRAM. It turns out bringing hundreds of transparent images and assigning them each to their own material is tough on the old graphics card even if the total size of the files isn't that large. I guess it's a good thing I got my GPU before the market went tits up.

The last things were the titles/logos. This was easy enough as it wasn't anything I hadn't done before. I just brainstormed a few ideas, downloaded a few fonts and assets, and threw something together.

The Security Room

I spent a weirdly large amount of time on this one, without any real good reason.

I was working on the materials for the servers behind her and for some reason I was really dead-set on having random alternating flashing lights. But I wanted to do it proceduraly. So I went down a whole geometry nodes rabbit hole to make it work. Was it worth it? Eh. You make some weird decisions in the middle of project slumps.

The CCTV and monitors in the background were just looping footage and I cobbled together the rest, like the wires and the flooring. Not much else to say about it but I think the thing as a whole turned out pretty nice.

Graphics

You could put this in editing but it fits here too. Due to the practically filmed nature of this whole animation, the dozens of different cameras required a bunch of different overlays and transitions to help further distinguish them and contextualize their narrative role. The hidden cameras needed to look different than the audience cameras, the audience cameras needed to look different from the main perspective cameras, etc.

So I gathered a ton of different UI and overlay assets and went to town on it in post. There was probably a lot more I could have done (at one point I had thought of little facial-recognition squares in some shots) but in an animation like this I didn't want to be overly intrusive.

It certainly presented a challenge from a watermarking perspective. I didn't want my name blasted into the corner of every shot, so I had to find creative ways to work it in to most of the angles. I think now you could call it more of a subliminal message than a watermark.

Desktop

This was the really fun part.

I had developed a technique for doing 2D animations in Blender back when I did my little animated Animation Challenge intro. It essentially consisted of just importing everything as 'slides' (mesh planes), overlaying them one-by-one as transparent emissives, and animating them.

So that was my approach with this and it pretty much went off without a hitch. I built almost the entire SombrerOS desktop and all of its windows from scratch in photoshop and it resulted in over 100 slides. Then I just imported them, organized them, and parented them to their correct partners.

And here's how the slide system looks in the viewport:

Pretty simple, really. Makes me want to take up 2D puppet animation. The dope sheet was a bit of a mess though:

Animation

As usual I try to go for fresh poses and animation obstacles with each project. Luckily this one wasn't quite as challenging as some past projects, but there were still some hitches.

Cloth Simulation

Thanks to my brutal, grueling cloth simulation trials in 2024 I was well equipped to deal with all of the dresses and jewelry this time around. I had to use just about every trick that I learned but it was all worth it. Plus, the dresses and lingerie in this project were built from scratch and optimized for simulation.

Still though, it was a relief anytime a character stripped down.

Soft Body Simulation

At this point I've developed so many techniques for simulating parts of the body that I rarely have to come up with a new solution when approaching something. However for this one, I needed a mesh cage that worked better for Ashe's dick. Usually I stick to hand-animating anything remotely rigid so I have the most amount of control, but I just wasn't getting the results i wanted here. So I formulated a spring-based system and honed it in until it was flexible enough to work well in many situations. Especially for the desk slamming scene.

Just about every other method had its chance to shine in this animation too. Hand-animated jiggles, soft-body edge target simulated jiggles, pressure-based mesh deform jiggles, wiggle add-on jiggles, shapekey jiggles, pure soft-body jiggles...

I still feel like I need to land on some better butt and thigh jiggles though. I have a few solutions right now, but they don't always feel like the right choice and I feel like I can do a lot better with them. I might have to focus on developing those techniques in the future.

Fluid Simulation

You don't get to this runtime without the expectation of a lot of money shots, especially not with this many perspectives.

We took a similar approach to all of the jizz this time as we did in previous animations, but were running into more trouble. The Widow/Tracer and Brigitte/Juno ones were fine, however the big foursome one yielded a lot of challenges.

It's not really my place to extrapolate on all of the technical difficulties as Morfium was the one simulating them, but the end result was that we were having trouble with consistency and motion blur. Motion blur had been a precarious process for all of the simulations, but it was especially problematic here. Since the simulations are ported from other programs as alembic files (.abc), Blender essentially reads them as a new mesh for each frame. Because of that, it needs additional data to understand the vectors and velocity of each mesh. We went through tons of re-exports trying to fix it, but in the end settled for something close. Still though it isn't quite right and can appear choppy in certain places.

I had also run into issues with the way I was shading and rendering them. I took the approach of rendering them in their own layer using Cycles so I could get proper volumetric density, but that approach hadn't worked as well as in the past. The environment and lighting set-ups were more complex, so the discrepancy between Cycles and Eevee was greater. Plus I was using a holdout masking approach which made rendering easier but might have affected the quality of the final product - even if it gave me more flexibility.

The main issue though was with Zarya's shots, which was an issue with getting the consistency right. After everything was rendered it really hadn't turned out how I'd liked, but by then we sort of needed to move on. I wanted to try and do a hand-animated version, but I decided to put that time into the Brigitte/Illari cumshot instead.

It was a bit of a late decision to add this in so I just did it all myself. It turned out alright with some obvious limitations to the mesh-based hand-animated approach.

Each of these shots could have gone better and - as is the way of fluid sim - there's tons of room for improvement here.

Poses

The practical aspect of the animation made pose and angle choice much more restrictive. When it's just two people in a dimly lit room like the previous animations, it's almost freeing. It's going to be a certain way and you get to focus on the other details. However at this scope and size, it can really tax the brain.

Each storyline had its own level of difficulty.

1. The bathroom was pretty simple. Illari was the observer for most of it so no matter what the camera had a lot of flexibility in its positioning. The simple environment made things even easier (aside from the mirror).

2. The foursome up in the VIP area had to deal with the spacial constraints of the lounge and the fact that every participant was active. So I didn't have the ability to go for interesting angles whenever I wanted.

3. Ashe and Widow were a combination of hidden camera and crowd perspective, which made their poses very specific. No wide overhead shots, no third-person shots, no real close-ups. This pushed their scenes into more of a POV style.

4. Tracer and Widow relied more on crowd perspective shots, which brought everything down to below eye level. I had some flexibility but I couldn't just tack another angle on there that looked cool. I had to justify and animate everything.

5. The security room was a singular constrained perspective of a webcam. So, the poses had to be built around the camera angle as opposed to the other way around. I had done something similar in the past, but this was still pretty refreshing.

At the same time I wanted to come up with positions that were at least somewhat fresh and interesting.

Editing

Girl's Night Out was meant to be more of a montage, so the whole approach was a lot different than my other big animations. A lot of the pacing decisions were made during the animation and editing process as opposed to being planned before-hand. This approach came with some benefits and drawbacks.

The nice part was that since I had the option of intercutting between scenes, I didn't have to worry about continuity or transitions nearly as much. I could add or remove ideas without having to worry about how the characters got there or where they were going next.

The downside was that pacing became more difficult when cutting between different perspectives. For one, each 'storyline' needed to have a comparable amount of material or the pacing could feel a bit lopsided. Secondly, the progression needed to make sense between each perspective. Having one perspective move on to hardcore while another was still in foreplay could make it feel too disjointed. So in a way, despite having to worry less about individual scenes, it made me worry more about the overall product. This meant that I had to make additions that I wasn't expecting, but at the very least they were easy to add.

Layout

This is a lot more boring/technical, but I thought I'd briefly cover it for those curious since I've never actually gone into this stuff before.

I usually stick to singular perspective stories, so the actual editing process isn't too difficult. You have an idea in your head, a couple of scenes, or a storyboard to go off of. Then you just need a hand full of sequences to assemble it, rough it out, and then put the final cut together.

Because of the montage nature of this project and the fact that so many creative decisions were being made in editing, things got more complex.

I relied a lot more on nested sequences and breaking each sequence down into its own 'storyline'. This made doing the sound effects and vocals a lot easier. Since I was essentially doing 'in universe' cuts, there was rarely a need to blend or overlay the sound between scenes. So I could afford to silo things off into their own sequences without much downside. Plus it made it easier to send the voice talent their cuts, and easier for them to dub over them, since they didn't need to see the full animation. It was a lot easier to organize all of the vocals I got back as well.

Then I just built it up from there. I had assemblages of all of the imported footage and did all of the overlay, grading, and basic dialogue and sound effects there. Then I pulled those into a rough cut to help solidify the story and figure out the music. Once I had all of the footage I pulled everything into a final cut where I could do some extra sound work, tweak some things, and figure out extra graphics and transitions.

Sound

The basic vocal work, dialogue, and sound effects were pretty simple this time around. The new challenge came with the music and all of the sound effects that came with it. I haven't worked much with music before, especially not in such a way that I have to make a sort of reverse music video.

For Girl's Night Out I didn't have one big track to lay everything out to. I had some songs in mind when I was putting the story together, but nothing concrete. So it was a process of finding the right music for the right moments, and figuring out how to lead from one to the next without any jarring changes.

I made pretty heavy use of risers and effects to help prevent any transitions that really took you out of it, and it helped a lot having parts of the story based in different parts of the club. Going from loud to muffled to reverberating can really help hide any cuts made to help loop the music or achieve a better tone. The only transition I wasn't that happy with was the cut from Sombra to Ashe and Widow at 5:27, but I didn't have much time to formulate a different idea. Maybe the fade out with Illari and Brigitte could have used some more work too.

Speaking of things I wasn't happy with, I think I could have spent a lot more time on the sound and mixing in general but I had held onto the project for far too long at that point and I didn't want to put all of that extra time in. It's an alright mix for headphones but things can really blend together for speakers or lower quality options. Some basic sounds and dialogue gets lost in the mix and there's some other stuff that made it to the front of the mix that probably didn't need to take up that space. It's a loud club so it's a real challenge trying to fit everything in without being too overwhelming, but I think it would have been doable given more time.

Conclusion

I hope this was elucidating for some of you out there. I say this pretty much every time I release a big project, but it's true: this is the longest thing I have made. I already went into just how much work and mental energy it takes to do something like this in my previous breakdown post so I'll save it here.

Though there was more freedom putting this one together, it was still a ton of different environments, a ton of different characters, a ton of different poses, etc. I think I racked up around 30 different blender project files. I'm hoping my next big animation is just a few big scenes I can work on, but I know I really can't help myself when I get an idea - even if it makes me pull my hair out later.

If you made it this far then thanks for reading and thanks for your support!