If you go for sfw realism, i think the best is midjourney v4 
Guides Create sexy art using VaM + Stable Diffusion
- Thread starter JaxZoa
- Start date
S
SPQR
Guest
i'll definitely try it, thank you!If you go for sfw realism, i think the best is midjourney v4
so much to explore
S
SPQR
Guest
|
|
no photoshop and few iterations, 25 or maybe 50. in-paint is the key for nsfw from what I can tell
pinosante
Well-known member
- Messages
- 390
- Reactions
- Points
- 63
wow, and what model/settings/prompt did you use?
S
SPQR
Guest
photograph, 25 year old caucasian woman, short messy bun blonde hair, sky background,beautiful, gorgeous makeup, red eyeshadow, black eyeline, blue eyes, portrait, pouty lips, kiss,naked,nude, posing
Seed: 3312451, Sampler: ddim, Inference Steps: 25, Guidance Scale: 7.4, Model: HassanBlend1.4, VAE: vae-ft-mse-840000-ema-pruned, Negative Prompt: cartoon, 3d, (disfigured), (bad art), (deformed), (poorly drawn), (extra limbs), (close up), strange colours, blurry, boring,sketch, lacklustre,{watermark}, ((tattoos)), ((tattoo)), Prompt Strength: 0.58, Fix Faces: GFPGANv1.3 640x960
Seed: 3312451, Sampler: ddim, Inference Steps: 25, Guidance Scale: 7.4, Model: HassanBlend1.4, VAE: vae-ft-mse-840000-ema-pruned, Negative Prompt: cartoon, 3d, (disfigured), (bad art), (deformed), (poorly drawn), (extra limbs), (close up), strange colours, blurry, boring,sketch, lacklustre,{watermark}, ((tattoos)), ((tattoo)), Prompt Strength: 0.58, Fix Faces: GFPGANv1.3 640x960
S
SPQR
Guest
haven't figured out fingers yet though, they're often messy. maybe it's just the number of iterations
pinosante
Well-known member
- Messages
- 390
- Reactions
- Points
- 63
Wow those “photos” look awesome. I’ll download the Hassan model.
The hassan blend model is cool, but the real magic is in merging and training your own models imo ! This is my HyperBabes_v1 Model, made from f222, hassan,anything v3 and a few others!
S
SPQR
Guest
those look great!
that makes sense but i'm lazy and don't have the gpus, i have a feeling it takes a long time to train it and there's also the question of licenses if you ever want to share the models. probably now it's wild west out there but it might get catty in that sense. i've seen watermark artefacts on some of the models, there's surely smart ways to force them & prove the model used licensed material to train
that makes sense but i'm lazy and don't have the gpus, i have a feeling it takes a long time to train it and there's also the question of licenses if you ever want to share the models. probably now it's wild west out there but it might get catty in that sense. i've seen watermark artefacts on some of the models, there's surely smart ways to force them & prove the model used licensed material to train
Merging models is almost instant! For exemple you could take the Hassan, and merge it with Anything with a ratio of 0.2, and it would slightly anime-ify your renders! I'll write a v2 of the guide soon including how to merge etc (it's really easy)
For the licensing i don't know, my goal is to create content using the model, not redistribute it
S
SPQR
Guest
that's great to hear, i'll definitely check it out, thank you!
- Messages
- 113
- Reactions
- Points
- 93
And my favourites:
I started with VAM models and drifted
Edit : Here is the site that helped me a lot for the prompt. You just need to register though~~
https://prompthero.com/stable-diffusion-prompts
S
SPQR
Guest
from my banter with Dragontales, one of his models
Last edited by a moderator:
S
SPQR
Guest
this is kinda wild on alien porn, i have a feeling that's gonna be big
i gave it a go on my halloween alien model and then also let the AI take more artistic liberties with it. the last one is one of the coolest sci-fi illustrations i've seen, alien space bar vibes
i gave it a go on my halloween alien model and then also let the AI take more artistic liberties with it. the last one is one of the coolest sci-fi illustrations i've seen, alien space bar vibes
- Messages
- 113
- Reactions
- Points
- 93
often you have an arm or a second body that appears, despite the negative prompts ... for my part I do not insist too much, 4 - 8 tries, if it goes in all directions, I change the prompt. In general I noticed that the more you are precise, the more it looks like nothing lol
I also recommend adding an artist name like:
My negative prompt :
ugly, deformed, deformity, disfigured, malformed, ugliness, blurry, disfigured, mutation, mutated, extra limb, bad anatomy, long body, cropped head, cropped face, anatomical nonsense, malformed hand, long neck, missing limb, floating limb, disconnected limb, lowres, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, dehydrated, duplicate, morbid, mutilated, out of frame, extra fingers, mutated hands, poorly drawn hands, poorly drawn face, bad anatomy, bad proportions, extra limbs, cloned face, gross proportions, malformed limbs, missing arms, missing legs, extra arms, extra legs, fused fingers, too many fingers, long neck, loli, lolicon
S
SPQR
Guest
not at all, usually more than 50% look good first try
maybe you're missing VAE: vae-ft-mse-840000-ema-pruned
hassanblend is the most realistic model i've seen so far but you should be careful with it as it tends to make some faces look questionably young. vam already sometimes goes there unintentionally, mixed with this too you can get in trouble. someone brought it up on the reddit release also the solution is to add mature to the prompts and lots of stuff to the negative prompt like kevinBurinn's above. even emphasize them like ((((anime))))
something i did for patreon yesterday, did 16, picked 8 for patreon post, then did 5 variations on one as an example
these were 5/5 variations of one other image that was generated. i don't have the gpu to run 200 lol
the dude's patreon is https://www.patreon.com/sd_hassan if you want newer version, i think he releases models first there
Last edited by a moderator:
Not using hassan in my current model, but even with the Anything v3 model, and a heavy negative prompt to avoid it, you still sometime have really disturbing generation... That's one of the real negative of this approach, and why i stick to anime style, don't want those images in my head
- Messages
- 113
- Reactions
- Points
- 93
@SPQR Thanks for share this !
So I found another model source, F222
I didn't think there were so many, here is the source:
There are also different tests, to see the renderings.
And here my test :
Here for the list of models :
So I found another model source, F222
I didn't think there were so many, here is the source:
There are also different tests, to see the renderings.
And here my test :
Here for the list of models :
Stable Diffusion Models
Contact Changelog Stable Diffusion Stable Diffusion v2.0 Stable Diffusion v1.5 [81761151] [a9263745] Stable Diffusion VAE ft-mse-original Stable Diffusion VAE ft-ema-original Stable Diffusion Inpainting v1.5 [3e16efc8] Stable Diffusion v1.4 [4af45990] [7460a6fa] [06c50424] Waifu Diffusion Waifu D...
rentry.org
S
SPQR
Guest
nice! the contrast looks so much better with that one
that makes sense. yeah hassanblend seems to gravitate a lot towards that if you give AI more freedom.Not using hassan in my current model, but even with the Anything v3 model, and a heavy negative prompt to avoid it, you still sometime have really disturbing generation... That's one of the real negative of this approach, and why i stick to anime style, don't want those images in my head
the best application for it so far for me is turning cgi to photorealistic and AR. I can't get to run it at 2k to see how it looks at great quality but i think stockphoto sites are gonna have quite the headache with this. It makes all those paid model photos kinda obsolete. You could pose a model anyway you want in any 3d software, vam is quite decent for that too in vr
Those AIs make so many things obsolete, i think it'll change the whole art industry, and that's just for the Text-to-image AIs! Sometimes it feel completely surreal!
I'll delay the v2 guide a little, still learning so much i'd have to update it every day! But now i see how bad the v1 guide is ahah x)
pinosante
Well-known member
- Messages
- 390
- Reactions
- Points
- 63
Can't wait!
.
Does anyone have a guide on how to install and then activate the models? This is all new to me and I do have SD running and working, but not sure where to put the models and then how to get them running. Any help would be awesome.
