making houndoom R34 AI porn

it took me 2 weeks to make this and i might have forgot some steps.

if anyone have some tips what have I could done better, please PM me on discord. im glad for any feedback.

this is the starting pic

(best quality, high quality:1.1), 1girl ,feral houndoom, muscular, slim, feral:1.8, flat chest:1.9 ,nipples, ,large claws, pussy ,anus , digitigrade:1.6,  sad, large paws,
, by hioshiru, by ecmajor:0.3 , (by kenket|by totesfleisch8|by thebigslick) <lora:sergals_v.1:0.3>, <lora:fluffyrock-quality-tags-v3.0-vpred>, slim long muscular legs, 
, abs, anorectic, anorexia, angry, teeth, saliva drip, 
BREAK 
masterpiece, best quality, ultra realistic, 4k, 2k, (intricate:1.1), (high detail:1.3), , sharp focus, , photorealistic, , , photorealism, absurd res, highly detailed photography,, vibrant, insanely detailed, hyperdetailed, , bed sheet, multiple light sources, daylight, high brightness, low contrast, bright,


Negative prompt: boring_e621_fluffyrock_v4,  dfc,  bwu,  aid210,breasts:1.5 , breasts:1.9, dark:1.5, hair:1.3, anthro,
Steps: 50, Sampler: DPM++ 2M Karras, CFG scale: 9, Seed: 197854390, Size: 1088x768, Model hash: 821628644e, Model: EasyFluffV11.2, VAE hash: e9ed949371, VAE: kl-f8-anime2.vae.pt, Lora hashes: "sergals_v.1: b800c7267f5f, fluffyrock-quality-tags-v3.0-vpred: 27c5fb166e9e", TI hashes: "boring_e621_fluffyrock_v4: f80ed3fee940, dfc: 21c6ae158a7e, bwu: 70e376c5cf1d, aid210: afa9c1d8e13c", Refiner: coconutFurryMix_coconutultimatereali.4.4GB [229744f05e], Refiner switch at: 0.12, Version: v1.6.0

after 10 retries:

i decided to censor the trainer, due to seemingly looking too young

(best quality, high quality:1.1), boy laying on the ground, scared,  open mouth, erection,

BREAK 
anime


Steps: 40, Sampler: DPM++ 2M Karras, CFG scale: 9, Seed: 884653745, Size: 1088x768, Model hash: 9d6644b6c3, Model: bb95AssmixFurryModel_v11Bb95v90-2.1GB, VAE hash: e9ed949371, VAE: vae-ft-mse-840000-ema-pruned.safetensors, Denoising strength: 0.61, Mask blur: 4, Version: v1.6.0

the whole crotch area is quite low, i would even call it droopy,

i redone just the butt part and had to manually move it up. AI always put anus and pussy too low.

(best quality, high quality:1.1), 1girl,solo,  ,feral houndoom, muscular, slim, feral:2, , detailed puffy canine pussy ,anus , 
 , ,, <lora:fluffyrock-quality-tags-v3.0-vpred>, muscular legs, 
, abs, anorectic, anorexia, dripping pussy,
BREAK 
masterpiece, best quality, ultra realistic, 4k, 2k, (intricate:1.1), (high detail:1.3), , sharp focus, , photorealistic, , , photorealism, absurd res, highly detailed photography,, vibrant, insanely detailed, hyperdetailed, , bed sheet, multiple light sources, daylight, high brightness, low contrast, bright,  

 <lora:exposure_control_v10:0.7>
Negative prompt: boring_e621_fluffyrock_v4,  dfc,  bwu,  aid210,breasts:1.5 , breasts:1.9, dark:1.5, hair:1.3,  anthro , humanization, 
Steps: 40, Sampler: DPM++ 2M Karras, CFG scale: 9, Seed: 2166523800, Size: 1016x848, Model hash: 9d6644b6c3, Model: bb95AssmixFurryModel_v11Bb95v90-2.1GB, VAE hash: e9ed949371, VAE: anythingKlF8Anime2VaeFtMse840000_klF8Anime2.safetensors, Denoising strength: 0.79, Mask blur: 0, Lora hashes: "fluffyrock-quality-tags-v3.0-vpred: 27c5fb166e9e, exposure_control_v10: 38e9e004f311", Version: v1.6.0

i enlarged the pic to 1808x1280 . i just did it in Photoshop. no fancy upscaling required, they are all bad anyways.

I P2P the enlarged 1808x1280 picture with tiled vae and tiled diffusion. i didnt use the controll net for more creativity.

(best quality, high quality:1.1), 1girl,solo,  ,feral houndoom, muscular, slim, feral:2, crotch boobs  , ,large claws, pussy ,anus , digitigrade:1.6, , large paws,
, by hioshiru, by ecmajor:0.3 , (by kenket|by totesfleisch8|by thebigslick) <lora:sergals_v.1:0.3>, <lora:fluffyrock-quality-tags-v3.0-vpred>, muscular legs, 
, abs, anorectic, anorexia, angry, teeth, saliva drip, 
BREAK 
masterpiece, best quality, ultra realistic, 4k, 2k, (intricate:1.1), (high detail:1.3), , sharp focus, , photorealistic, , , photorealism, absurd res, highly detailed photography,, vibrant, insanely detailed, hyperdetailed, , bed sheet, multiple light sources, daylight, high brightness, low contrast, bright,  


 <lora:exposure_control_v10:0.7>
Negative prompt: boring_e621_fluffyrock_v4,  dfc,  bwu,  aid210,breasts:1.5 , breasts:1.9, dark:1.5, hair:1.3,  anthro , humanization, 
Steps: 40, Sampler: DPM++ 2M Karras, CFG scale: 7, Seed: 197854390, Size: 1813x1280, Model hash: 9d6644b6c3, Model: bb95AssmixFurryModel_v11Bb95v90-2.1GB, VAE hash: e9ed949371, VAE: anythingKlF8Anime2VaeFtMse840000_klF8Anime2.safetensors, Denoising strength: 0.7, Tiled Diffusion: {"Method": "MultiDiffusion", "Tile tile width": 96, "Tile tile height": 96, "Tile Overlap": 48, "Tile batch size": 4, "Keep input size": true}, ControlNet 0: "Module: tile_resample, Model: control_v11f1e_sd15_tile [a371b31b], Weight: 1, Resize Mode: Crop and Resize, Low Vram: True, Threshold A: 1, Guidance Start: 0, Guidance End: 1, Pixel Perfect: False, Control Mode: ControlNet is more important", Lora hashes: "sergals_v.1: b800c7267f5f, fluffyrock-quality-tags-v3.0-vpred: 27c5fb166e9e, exposure_control_v10: 38e9e004f311", Version: v1.6.0

not only the trainer released the houndoom, it seems hes been busy summoning eldritch entities as well.

i used both pictures...

i cut out the butt, enlarged it, and put to P2P. you want to enlarge it as large resolution as your GPU can handle without til

ed diffusion.

the higher resolution, the more details you can get.

i got a qestion to to put it back together so:

select the area
go to paths
selection to path.

now i have nice blue box at the area i cut out and its super easy to put it back. even if the picture is a different size.

(this program is gimp, but its practically same as photoshop)

i went with this one... for some time-being

 
(best quality, high quality:1.1), 1girl,solo,  ,feral houndoom, muscular, veiny muscles,puffy,canine pussy:2  , pussy juice, anatomically correct genitalia,
, by narse:0.3, by ecmajor:0.3 ,, ) <lora:sergals_v.1:0.2>, <lora:fluffyrock-quality-tags-v3.0-vpred>, muscular legs, 

BREAK 
masterpiece, best quality, ultra realistic, 4k, 2k, (intricate:1.1), (high detail:1.3), , sharp focus, , photorealistic, , , photorealism, absurd res, highly detailed photography,, vibrant, insanely detailed, hyperdetailed, , bed sheet, multiple light sources, daylight, high brightness, low contrast, bright,  





Negative prompt: boring_e621_fluffyrock_v4,  dfc,  bwu,  aid210,breasts:1.5 , breasts:1.9, dark:1.5, hair:1.3,  anthro , humanization, 
Steps: 40, Sampler: DPM++ 2M Karras, CFG scale: 9, Seed: 3514909258, Size: 744x1064, Model hash: 229744f05e, Model: coconutFurryMix_coconutultimatereali.4.4GB, VAE hash: e9ed949371, VAE: anythingKlF8Anime2VaeFtMse840000_klF8Anime2.safetensors, Denoising strength: 0.6, Mask blur: 0, Lora hashes: "sergals_v.1: b800c7267f5f, fluffyrock-quality-tags-v3.0-vpred: 27c5fb166e9e", Version: v1.6.0

or it may be this one.

and i used some parts from this one later on:

i fixed the horn

i liked this one, but of course it don't really fit so far.

by dr. comet,  by bonifasko,  by karabiner,  by wolfy nail,  by tom fischbach,  by hioshiru,  by totesfleisch8,
houndoom, teeth, angry, furious, 

 <lora:Loraeyes_V1:1>,
,looking down, , dripping saliva, 
Negative prompt: bwu, dfc, ubbp, updn, monochrome, 
Steps: 40, Sampler: DPM++ 2M Karras, CFG scale: 9, Seed: 3332283390, Size: 512x512, Model hash: 821628644e, Model: EasyFluffV11.2, VAE hash: e9ed949371, VAE: anythingKlF8Anime2VaeFtMse840000_klF8Anime2.safetensors, Denoising strength: 0.6, Lora hashes: "Loraeyes_V1: da8286d2ab99", TI hashes: "bwu: 70e376c5cf1d, dfc: 21c6ae158a7e, ubbp: 047acf26d29c, updn: b4ae8ca1b247", Version: v1.6.0

those might be similar but i took something i like from each image:

this is how one layer of the merge looks like:

and again, from this point on, with this amount of detail. im not working with the AI, i battle against it.

i had to manually smudge all inconsistencies this one and the the next picture as well.

im just unable to create more detail. this is the absolute maximum im able to do.

ive tried like 10 different models including SDXL with no improvement, i got either nonsense or lesser detail.

you should whiten out the eye for better eye generation. but even then its hard to get something nice. even with loras they are just not the best, so i did the eye manually.

damn, it cant do even smoke without being guided


teeth, ((((smoke out of nose,exhaling smoke,)))), saliva, 
BREAK 
masterpiece, best quality, ultra realistic, 4k, 2k, (intricate:1.1), (high detail:1.3), , sharp focus, , photorealistic, , , photorealism, absurd res, highly detailed photography,, vibrant, insanely detailed, hyperdetailed, , , multiple light sources, daylight, high brightness, low contrast, bright,  realism,
Steps: 40, Sampler: DPM++ 2M Karras, CFG scale: 10, Seed: 2924491663, Size: 894x1013, Model hash: 2bb2c42192, Model: bb95FurryMix_v120, VAE hash: e9ed949371, VAE: vae-ft-mse-840000-ema-pruned.safetensors, Denoising strength: 0.54, Mask blur: 0, Version: v1.6.0

i clone stamped the smoke, and the AI finally took over from here