Make sure you're using a convergent, deterministic sampler (i.e. don't use one with "a" or "SDE" in the name). Twins and clones carry identical art seeds, just like with WebGL, so as long as you don't diverge their prompts/poses too much, don't reseed, and use a deterministic sampler, they'll look pretty similar. Same thing is true for attitude changes and most clothing changes.
This is unfortunately not sufficient, because I let my obedient slaves choose their own clothing. In combination with minor variations in height, breast size, obedience and trust, this is usually sufficient to cause twins and clones to look significantly different with AI, while they still look like twins with WebGL or vector art.
Neither WebGL nor AI art can do anything special with related slaves that aren't twins or clones; if you notice similarities in WebGL it's just because they have similar physical parameters and WebGL doesn't have that much variation even between unrelated slaves of the same ethnicity.
I agree, but in this case I consider the lack of variation in WebGL or vector art to be a feature, not a bug.
This allows me to easily recognize slaves that are part of the same family (siblings or parents/children), usually because they share the same skin tone, hair color, and main facial features. WebGL and vector art only support a limited number of combinations of those, but this is a good thing because it makes it easier to recognize the slaves and the similarities between related slaves.
Here is an example of a (futanari) mother with three different sets of clothes. You can see immediately that these are all pictures of the same character:
And here are three of her daughters. They share the same skin tone and hair color. Although they have very different haircuts and clothes, you can see that they share some facial features with their mother, so it is easy to recognize them as being part of the same family:
On the other hand, if I pick another slave that is not related to them but that also has a fair skin tone, and if I dress that slave in the same schoolgirl outfit and the same haircut, then the faces (and body proportions, etc.) are still sufficiently different that it is easy to guess that they are not related:
With AI, these similarities and differences are drowned in the noise because there is just too much variability. AI-generated images can look incredibly good but the excessive variability in physical traits and attitudes is working against the playability aspect. I love the photorealistic portraits that can be generated with Stable Diffusion, but I still prefer WebGL for the playability because it is easier for me to recognize the slaves.
I think I have 'the average computer' but will have to revisit my views when I get my new one built. In general I rank WebGL lower than you do in almost every field as I find it exceedingly heavy compared to AI just being slow. AI I had to run overnight but I can have images for all my slaves that look passable while WebGL I either have to turn the per slave options way down to the point it looks like garbage or disable all settings for mass image display.
I am probably lucky, but WebGL is just working fine for me on my laptop. The CPU is an Intel Core i5-1145G7, which is about three years old. The graphics are handled by the built-in Intel Irix Xe Graphics (no Nvidia nor AMD GPU). I can easily scroll through a list of more than a hundred slaves rendered with WebGL and it works fine. Here is a cropped screenshot from one of my current games (an arcology populated with pairs of twins in incestuous relationships) that has 135 slaves, most of them in the penthouse:
On the other hand, the AI image generation with Stable Diffusion does not seem to benefit from any significant acceleration on my laptop, so it is unfortunately very slow.
user/Bluecoffee93/posts
user/Bluecoffee93/posts
