Hi there, been posting for quite a while now. SexyFatGirlRenders

[SexyFatGirlRenders]

On the issue of skintone I’m just wondering if the subtle change in lighting that I currently have for side head images, is something the player would understand to be due to the fact that that is some sort of special closeup picture of her face.

I mean I could literally just make the side head image a zoom and crop of the normal sprite. That would literally make the skintone match perfectly pixel for pixel.

The reason I didn’t do that is that this arrangement is higher quality. The problem is that high quality lighting for a body and high quality lighting for a face ended up resulting in two very different skintones. (Although like I said, the main problem is the differences that arise from trying to highlight pale skintones versus darker skintones.) Of course this girl’s entire design is flawed. I was writing a hispanic or latinX woman. This I think is maybe more of a middle eastern skintone? I’m just trying to get things to fit.

It’s still pretty early on. So this stuff is still up in the air.

I redid this girl’s smile side head image with her new hair.

Which hair do people prefer? The more boring hair is simulated to lay on her. Whereas the other hair is basically just a static prop.

On the issue of the deadline, let’s say the heat death of the universe? Using sprites like this is maybe 1000 times easier/faster.

EDIT: Also to clarify, side head images pre-date my desire to use sprites. In other words. In a fully rendered scene. There would still be a side head image that might change to denote mood changes. Changes too small/subtle to effect the full scene render. Side head images are keyed to automatically change to match the mood of the say command. But now that I’m planning to use way more sprites than full scene renders… I’m beginning to wonder if the side head image is now flawed as concept? Because well… now that I have sprites. I can key the sprite to the say command and the mood based image attributes in the exact same way as the side head image.

TLDR: Maybe side head images are only useful in games with full scene renders. I’m just not aware of this because games on this website do not typically use side head images. In that case maybe I should simply turn side head images off. Then I wouldn’t have to worry about them clashing with the sprite. But I worry that full body sprites won’t show as much emotional detail as the side head image. A decent amount of effort has gone into fine tuning her exact facial expression in the side head image.

[SexyFatGirlRenders]

If you feel like the two skin tones showcased for this tan girl are too far apart to make sense… Then which do you prefer? Because I can always swap the lighting rig and tone mapping between them.

This whole skin tone issue is one that I’m also quite mystified how no one else runs into it. Because this minor a change can just be due to the difference between a set of daylight lightbulbs versus the sun. To clarify, both scenes have the same skin textures, the same number of lights, almost every detail is identical except for a minor shift in tone mapping and the temperature of the light bulbs.

[Krodmandoon]

@SexyFatGirlRenders

Which hair do people prefer? The more boring hair is simulated to lay on her. Whereas the other hair is basically just a static prop.

The first render image with the “more boring” hairstyle looks better to me personally.

@SexyFatGirlRenders

If you feel like the two skin tones showcased for this tan girl are too far apart to make sense… Then which do you prefer? Because I can always swap the lighting rig and tone mapping between them.

I would say the head render and the sprites both make sense. But looking at the sprite it’s like she’s inside of a football stadium being super lite from multiple angles which of course distorts her skin tone due to how cool the color temp is. It’s not a huge deal as I like lighting in a lot of the renders and the renders themselves look good but obviously the lighting doesn’t match the background image or what I would expect a dinner to have for lighting. And they stand out more because of the difference between the head and body sprites.

I would say going with warmer less “daylight” white color temp for the lights would help with the disparity between the two.

[SexyFatGirlRenders]

It’s not a diner, it’s a cafeteria in an office complex, so the lighting is harsh?

This is the diner image…

And here’s some other random maybe prototype background images

a bunch of them are over the filesize limit for the forum, so, https://gofile.io/d/YKPj4D

TLDR: You’re making me think I now need to double my number of sprites and make an indoor and outdoor lighting variant. The harsh cold light would be indoors and the warm soft light would be outdoors…

EDIT: Also, uh yeah, for indoor scenes I typically remove the ceiling and one wall and just use a custom lighting rig. Using the actual lights in a scene can be problematically dark. I think of my scenes more like hollywood studio stages than reality. But yeah I have between these two skin tones intentionally shifted the temperature of the custom rig’s bulbs from like 9000k to 4000k roughly. Just off memory, not at the desk right now. I felt like this change was justified at the time. But that was for a different skin. Also, in some regards, I’m washing this lady’s skin out on purpose. Since I want her to be Hispanic skin tone, not Egyptian goddess skin tone.

[dingotush]

@SexyFatGirlRenders

This is the diner image…

The depth of field is so tight on this it makes me think of macro photography of doll’s house furniture - maybe ease off a tad?

On the side-images, my gut reaction is to use a consistent warmer lighting to enhance the face and the expression even if it doesn’t match the scene itself. But it’s going to be a suck-it-and-see thing really.

[SexyFatGirlRenders]

Yeah the blurry image is actually in a pile of rejects. But it’s a reject render. The scene itself is still saved and if I need a diner I plan to revisit it potentially. And/or make it a pizzeria, or you know repurpose the scene for some purpose. Also the side head image is already using warmer lighting, that enhances the face, and is consistent, at least I tried to make the render nearly identical between the happy and angry faces.

In other words… are you agreeing that having the side head image as they are currently set up is good? Or are you saying it should be EVEN WARMER?

[SexyFatGirlRenders]

Atm the left hand side is a tooltip (appears on mouseover). The right hand side is the previously purchased item. The center is the real and normal renpy choice menu buttons.

Still a bit of a UI prototype, but could use some feedback.

[BadIdea4]

Idea: Keep the picture, but put it on the left and make it show the last-hovered item, like a tooltip. On the right, show a written list of what’s ordered so far.

[SexyFatGirlRenders]

So, the overly blurry diner image. The reasoning behind the blur in this case… prevent disconnects in the reality of the game. It is IMO far too much work to redesign these text menu type precanned assets to actually match the in-game choices for the player in terms of what they can actually order. Hence blurring it.

So which would make a better background?

edit: The quick and simple gaussian blur in gimp I think is a more targeted way to destroy this disconnect than using camera depth of field like in the rejected diner render. I think the better time to use depth of field would be for full scene renders, to make the female figure pop. ESPECIALLY if she’s not following the thirds guide rule. Although the main reason to do such a thing would be in scenarios where you move the focus of the subject of the render, without moving the camera. This would be a great way to do a striptease or wardrobe malfunction. But… it would take me eons to get right, so I may never end up using this trick. edit2: I’m more likely to do it for a strip tease than a wardrobe function. I’ve seen it used for stripteases. A striptease is very easy. I’ve spent a lot of time working on wardrobe malfunction style effects and only been mildly satisfied. I might try combining a striptease style scene composition with wardrobe malfunction style writing. That could work. Maybe. I’ll have to see.

[SexyFatGirlRenders]

I’m not at all happy with her expression or pose or even really her likeness to the intended character. I do really love the outfit though. The outfit is very on point. I might totally redo this scene with a generic genesis9 girl instead of this genesis8 lookalike.

That’s another advantage of sprites. I don’t have to live with 100s of failures. Also making backgrounds and sprites isn’t particularly a waste of time. I can load their scenes. resave them as scene subsets. Load them into the same scene that way. Then tweak the positioning and poses and such in order to create full renders.

Of course this workflow assumes I can live with a game where the player is only granted full renders like 10% of the time. Major stages and/or non-repeatable events. That sort of thing. Again inspired by thicker treat.

EDIT: really it’s her face that I entirely hate.

[SexyFatGirlRenders]

The p in coffee shop dangles down over the image instead of staying up on the border label. Is this good or bad? I did this on purpose. I can use a larger font size if the text is allowed to dangle like this. But maybe I’m the only one who is this blind and the average gamer prefers smaller text that fits where it belongs?

[SexyFatGirlRenders]

Oh, btw another advantage of sprites. With a 4070ti with 12gb of vram, a fully rendered scene with two femalenpcs and a detailed backdrop enviroment will make me run out of vram entirely and force me to hide/delete objects/textures not in the view of the camera.

If I owned like a 3090 or something instead and had 24gb of vram fully rendered scenes would be a lot easier to fit inside the vram limit.

Even with just one femaleNPC in my chelsea renders on deviantart I had to delete like half the diner to fit inside the vram limitations when I was going above 1080p renders.

ATM I’m doing 900p sprites, but I could easily do 2160p sprites if I wanted in terms of VRAM. Maybe I should move to 1800p sprites though? That’d let me oversample harder (I’m doing at2 but could do at4). Oversampling means that ren’py can do a zoom in effect without losing quality. Oversampling also provides better support for monitors above 1080p resolution, which is becoming more common.

Oh, also also, a 900p sprite can render in like 5minutes but a fully rendered scene even at 900p could easily be 1-6hours.

[SexyFatGirlRenders]

Color balance preferences. I just realized maybe my monitor settings are wonked up. Maybe I’ve been seeing red and blue tones a bit “wrong”. So, feel free to weigh in.

Color balance I came up with when trying to make her look like I wanted her to look. With a heavy emphasis on her being indoors in an office with those office building fluorescent lights:

Pure zero, BUT in bright sunshine, specifically at a real beach called spiaggia di mondello:

Default color balance I’ve used in 95% of my renders (before I started working on this character design, that is):

[BadIdea4]

I hear it works well to find photos from the environment in question, drop the sprite in, and see if the color balance matches / looks plausible.

There’s also the factor that IRL we take a bit to move from one lighting environment to another, giving our eyes time to adjust, but in a game it can be as quick as one frame to the next. In that case, it might be better to artificially adjust all color balances toward center so players aren’t wondering “why is everything bluish all of a sudden?”. Kind of a gameplay trumps realism thing.

Or maybe people take it in stride anyway. I never ran the experiment.

[jcc309]

I agree. I think with renders, I would default more towards “easier and nicer to see” than hyper realism. That doesn’t mean you can’t or shouldn’t change it up a little bit depending on the scene, but I would rather high quality renders that are easy to visualize than complete realism.

[dingotush]

I’ll prefix all of this by pointing out that I’m protoanomalous (not good at seeing red), so I’ll let that colour your judgement of anything that follows…

People use colour saturation, contrast, tone, and level of detail as cues to the depth of an image, especially a flat one on a screen. As we usually focus on the interesting foreground things in the distance are more blurred, which is why even a limited blur will set something into the distance. As we are used to atmospheric haze we also interpret lower contrast, lower saturation, and in general blue tones as distant.

Since the say box image is notionally part of the “HUD” and therefore “closer”, it makes sense for it to be warmer than the main image, if only slightly.

You can also play with the emotional response to colours: generations of living by firelight and incandescent light leave us associating “warm/yellow” colours with being happy and cozy, and “cold/blue” tones with discomfort. This is bizarely true even when it’s technically not correct: a midday sun is hotter and has more blue in it than the “golden hours” after all, and an oxy-acetelene flame is hotter than a pile of burning wood. Yet, it’s the burning wood we’ll gather around.

The other thing to consider in a render is the colour and angle of the lights. While it might be tempting to put every light source in as it would be and let the system do its thing, there are some potential things to try from stage lighting to emphasise the characters/actors. The main one is having a warmer and cooler light coming from two different sides so they don’t look “flat”, even if this isn’t “natural”.

[SexyFatGirlRenders]

I did try experimenting with blue light red light motif to generate a bit of an effect but I found it didn’t go super well…

but that might be because this is a closeup showing off eyeballs and a regular scene with blue light red light tones on the lighting rig around the central character would pop better?

I’ve been trying to make sprites as “neutral” as possible. Which wouldn’t be my goal if I was doing a full scene render. Or at least I’d hope it wouldn’t be.

TLDR: The main reason I flip flop from all 9k temperature to all 4k temperature on a lighting rig is mostly laziness, a lot of them actually default to like 6k for some lights and 4.5k for other lights to try to generate the sort of effect you’re describing… hmm, a good read though either way.

[dingotush]

@SexyFatGirlRenders

I did try experimenting with blue light red light motif to generate a bit of an effect but I found it didn’t go super well…

Thats’s … a lot.

So for warm colours you might use these gels:

• 007: Pale Yellow
• 009: Pale Amber Gold
• 013: Straw Tint

For colder colours:

• 003: Lavender Tint (basically gets you to “white” in front of incandescent lights)
• 063: Pale Blue
• 117: Steel Blue

Gel filters are subractive: they remove/reduce wavelengths to give a coloured light. If you hover a swatch it will suggest uses, and if you click on one of these it will take you to a transmission chart so you can see what it does physically. Also when used practically they’re placed out of focus, so you can double up or use a half-sheet to change the intensity. Colour temp alone doesn’t really describe what’s going on here.

When the light hits the actor that’s additive: so you can sill have a neutral effect overall from the front, but surfaces angled to the side will take on more of the colour from that side, making the 3 dimensional nature pop. Having any difference in the colours helps. This is less about realism and more about impression and emotion.

[SexyFatGirlRenders]

Okay so this is:

Front right pure white key light average strength.
Rear left back light pure white weak strength.
Instead of a fill light I like to use an hdri image including a sun, beach shot, angled for front left but fills entire scene with light, just it’s mostly aimed as if the sun itself were on the front left, average strength, not tinted.

For this new render:
The front right key light was color shifted to a bit of a reddish straw, sort of like 009, but a bit closer to 206, strength increased to strong, and temperature also dropped from white white to warm light bulb.
The rear left light was tinted as 061 and strength increased to average.
The environment slash sun was cut in strength by roughly half. (I think I cut it a bit too much, but I wanted to offset the increase in strength on the other two light sources.) Also tinted towards 061.

I think your suggestion was intended for an indoor scene without sunlight and without an hdri and possibly for a portrait with a specific key light and fill light in red and blue tints respectively?

Sidenote: the white balance in this scene is about like 004 which is a negative removal effect that reduces the redness in the scene, this was chosen purely because it happened to be true for render1, so I kept it roughly the same for render2, this choice was originally made to get the skintone I wanted because I was too lazy to alter the skin texture itself.

Sidesidenote: The red and blue lighting in the “bad example” was intentionally hamfisted because I wanted to be sure I noticed the effect, it could easily have been made more subtle, I just found the effect so bad that I didn’t ever bother to try to fix it.

TLDR: I think I understand your instructions, but applying them in the halfassed manner I did in this example didn’t produce the desired effect precisely. Then again, maybe instead of using this idea for a sprite or a scene it should be used on sidehead closeup images.

Edit: The 3d effect you’re referring to, I think might be functioning on the shoe she is wearing on the foot that is on the left of the image. Even though that detail was meant as an afterthought.

[dingotush]

This is doing more or less what I expected.

In the second image the dimensionality on her skin is improved: the brow furrow, cheek hollow, neck strap muscles, and collarbone are better defined. All this makes the expression overall easier to read. Bonus side effects are glorious hair, more natural looking skin, and more definition on her legs. What’s been lost is the pinstriping on her outfit; but I’d say that’s less important and a good compromise. That’s what I see anyway.

Obviously in a stage setting being able to read the actor’s emotions/expressions is key (while not losing them against the set), so a degree of over emphasis is used (also with eye and lip makeup particularly). You can work this for indoor, outdoor, and any time of day. You just need to get the overall colour right (for the scene), and ensure there’s more difference in colour from either side than you might think “reasonable”.

It’s also trivially easy to get the overall effect right when your just moving sliders and having it play out in real time and not having to wait for a render, but changing colours is much harder involving ladders and scaffolding! The rig would be set up for all the scenes, be it daylight, indoor, dawn/dusk, and nighttime, so there are always colours to play with - just never the same from both sides.

Anyway, it’s something to play with. A way to draw attention to the things you want to emphasise over a commitment to realism.