CivArchive
    Preview undefined

    Just another merge of some popular models, that I think turned out good.

    It's somewhere between 2d and realistic. (swings back and forth with every version) The merge is a bit biased towards nudity, and big breasts but is still pretty versatile. (see example images)

    The older furry/futa variant is more 2d-2.5d than the other versions. Version 2.0b, is more 3d/realistic.

    For best creativity/flexibility, I suggest trying v1.8-1.9 if you can deal with the inconsistent/bad hands, otherwise use v2.0b which are more grounded and stable but less flexible.

    There is no VAE baked in versions below 1.8, so I suggest using either vae-ft-mse-840000-ema-pruned.vae.ckpt or kl-f8-anime2.ckpt with the merge.

    Try to keep the negative prompt to a minimum. ("standard" negs like worst quality, low quality etc. seem to have almost no impact in v2.0b, comparison here: https://civarchive.com/posts/892866 )

    Something like "(nude:1.3)" is enough normally.

    Strong negs like (worst quality, low quality:2) suck out any variety and flavor.

    The merge works well with either clip skip 1 or 2.

    Changelog:

    Update 2.0b Futa (with baked vae and pruned)

    Mixed in two futa loras that work somewhat well and allow for different dick sizes. (Futanari_V1 and tgirls_V4)
    Use simply "penis" for random dicks, or prompt "small erect penis" with "big penis" in negative or the other way around to get somewhat consistent results. (Should also work with flaccid.) A difference between "big penis" and "huge penis" is there, but small... lol

    The stability depends heavily on angle/pose and complexity, so try to keep the prompts simple.

    Update 2.0b baked vae and pruned

    mixed 2.0 with the older 1.8 and added bits of the following models:

    • 0001softrealistic_v154

    • hardcoreHentai_v12BakedVAE

    • pornmasterPro_newv1

    • realDream_8

    • realisticVaginasGod_godpussy1V2 (Lora)

    The result looks almost the same as 2.0 but seems far more stable, better and more consistent hands. (Hands still fuck up sometimes, but far less than with earlier versions.)

    A bit more lewd though, biased towards nudity, but a "nude, nipples" in neg helps to fix that. If there are still too many naked breasts try to word it differently, like instead of breasts use cleavage etc. Or you can skip the prompt of breasts altogether, the model isn't conservative with its usage of cleavage/curves just by prompting woman or female. (like in "female thor, female hulk etc. but the effect depends on how "hard" the armor of the character is, ironman armor will need a little prompt "boost")

    Update 2.0 baked vae

    • reduced the lazymixRealAmateur part and added a bit of analogMadness

    • generally more stable

    • better hands, faces, feet etc with 2.5d and 3d images (better, not perfect... and nails are still a bit iffy)

    • a bit more biased towards nudity

    • i recommend neg (worst quality, low quality, nude:1.2) as base and then adjust as needed

    • model leans now a bit more towards realism, fantastical stuff needs a bit more weight, but should still work

    • for 2d comic style stuff the model still works, but 1.8 or 1.9 are better

    Update 1.9 baked vae

    • added lazymixRealAmateur_v30b

    • more of a side grade than an upgrade

    • more consistent/better hands/feet/eyes (hands and feet are still often inconsistent, really depends on the prompt)

    • less stylized and more realistic, but also less flexible

    • more biased towards nudity

    Update 1.8 baked vae

    • just baked in vae-ft-mse-840000-ema-pruned.vae for easier use with generation service

    Added inpainting version of 1.8

    Update 1.8

    • Added models:

      • beenYouLite l15

      • realcartoon3d v4

    • straight up upgrade from 1.7, maybe a bit worse hands...

    • better details, better contrast, better age range (can finally do older skin)

    • a bit biased towards muscles, just adding "muscles" to the neg should be enough

    Update 1.7-futa (just a small spin-off)

    • merged in futanariFactor_alphaV10 and FutaVeiny v6

    • can now do somewhat consistent futa dicks and bulges, see examples

    • like the furry spin-off, a bit more 2d than the normal model

    • add to prompt something like futanari, penis, erection or bulge, depending on what you want, see examples

    Update 1.7

    • added Cyberrealistic v3.2 to v1.6 furry b

    • more stylized than v1.5 but also more detailed and versatile. (and bigger tits…)

    • better for fantasy, less biased towards nudity (unless prompted)

    • sometimes adds cat ears with fantasy words… (remnant from the furry model)

    • furry can still be forced with enough weight

    • age is harder to change than in v1.5, but still possible (negative embeddings tend to normalize age)

    • Negative can be reduced to "(worst quality:1.1), (verybadimagenegative_v1.3:0.9)" and get good results

    • try to use the least number of negative embeddings to get your results, they tend to suck out variety

    Update 1.6 Furry (next version will most likely be non-furry again, just wanted to give it a try :3 )

    • added Indigo Furry mix v40

    • two versions: (see example images)

      • Furry_A: more stable, but less details

      • Furry_B: more details, but less stable

    • prompt should include: (furry, anthro $race/animal woman:1.3)
      (other tags like bodyfur, snout etc. help, see example images)

    • biased towards muscles and human skin, adding the following negatives helps: (abs, athletic, muscles, muscular, human, human skin, skin:1.2)

    • can do non-furry, just exclude the furry words (faces look kind of the same, though. but with tags like wrinkles, makeup, freckles etc. variety is still possible)

    Update 1.5

    • added model darksun

    • added LoRA "Add More Details"

    • resulting in increased contrast and details

    • Negative should include: (worst quality:1.0), (low quality:1.0), (bad-picture-chill-75v:0.9), bad-hands-5

    Added inpainting version of 1.4b

    Update 1.4b

    • jam v1.4 + some nextphoto v1.0

    • style didn't change much, but faces are more realistic and the merge is less biased towards nudity or blonde hair

    Update 1.4

    • Replaced fantasticmix with schExcelsior. The model was getting too realistic, now it's more even between 2.5d and realistic.

    • Merged realisticVaginasGod_godpussy1V2 into the model.

    • Breasts/Nipples are more consistent now, pussy is still a bit random.

    • Can't do old people -.- trying to fix that with then next version.

    Update 1.2:

    • Added fantasticmix_v50 to the merge, now it should be a bit more flexible with fantasy/sci-fi stuff.

    • Accessories, like necklaces, also are seemingly more consistent.

    Updated models in v1.1:

    • uberRealisticPornMerge_urpmv13

    • henmixReal_v30

    • cyberrealistic_v21

    • revAnimated_v122

    • deliberate_v2

    Description

    inpainting version of 1.8 pruned

    FAQ

    Comments (16)

    nbwgundaop954Jul 28, 2023
    CivitAI

    my images are to noisy/deformed while using this model

    Purfecter
    Author
    Jul 29, 2023

    what are your settings? prompt, neg etc.

    D00derinoJul 29, 2023
    CivitAI

    Great model (lately a bit too dark and gritty IMHO, but it's a matter of taste).

    I'm curious about something: when I try to reproduce your images, the result is like 99% the same, but zooming in and comparing with the original, every detail is slightly different, and my image is also more saturated. Any idea why?

    For example: https://civitai.com/posts/442563

    If it helps, I'm using sdp-no-mem and vae-ft-mse-840000-ema-pruned.safetensors

    Purfecter
    Author
    Jul 29, 2023· 2 reactions

    That's because you most likely (and everyone else) uses xformers to speed up generation. But it adds a bit of randomness, so even with the same seed, prompt etc., the images will be slightly different. (sdp-no-mem is also non-deterministic, so it adds randomness) Honestly, no idea why the saturation is different. ^^

    Purfecter
    Author
    Aug 18, 2023· 1 reaction

    The model can also do bright and colorful, if prompted for it, see: https://civitai.com/posts/513126

    1996malmoa714Aug 5, 2023
    CivitAI

    When I try to mimick the brazilian picture by copy generating the data it turns out all blurry and the lights are very yellow. I downloaded the same Hires upscaler(4x_NMK supercale) and the bad-picture-chill. Only difference is that my model is "jamJustAnotherMerge_v18Pruned" instead of "jam_v18-pruned".


    Purfecter
    Author
    Aug 5, 2023· 1 reaction

    The model is the same, it just gets renamed when I upload it. Otherwise, not really enough info to pinpoint anything. With the upscale keep in mind that the 10 steps are enough for 0.35 denoise, for more you would probably want more steps, or it might get blurry. Its also important which tool/UI you are using, A1111 and ComfyUI for example, will produce slightly different images, since the settings are never 100% the “same”.

    1996malmoa714Aug 5, 2023· 1 reaction

    @Purfecter Thanks! Your right when i changed the High res and steps it works perfectly.

    ThatPurpleOneAug 11, 2023· 3 reactions
    CivitAI

    This model is absolutely excellent! In 95% of cases, I use this model as my starter txt2img model - and depending on taste and mood, as my final upscale img2img model as well, though its initial outputs mix very well with other models.

    What sets it apart for me is that it responds very well to prompts, and it seems to have a very high variety of images it was trained one, with all its merges. NSFW works just as good as SFW, and if you use one of the Detail LoRAs for both the txt2img and the img2img upscale, you almost get a very artsy look.

    11/10, best model for my own tastes. Can't wait to see what the next versions bring.

    Name_Already_TakenAug 16, 2023
    CivitAI

    The most usefull download i made on this site.

    -It understands just about anything you throw at it (including anime&realisitc loras).

    -The cartoonish look has gotten stronger over the versions. It s a nice look in its own right, ...

    - .... but I tend to switch to realsitc models in img2img, which works like a charm.

    Hence while it s the most used thing for me in SD, it s showing up the least in the end results. However it makes getting to the desired end result so much easier.

    Great Work

    itsthehandAug 16, 2023
    CivitAI

    I have seen what this model can do and I love it but when I try to use it, it makes everything noisy and high contrast and over saturated. The image looks amazing as it is generating but the last 4% it suddenly adds some filter that over saturates and such. Wonder if anyone would be able to help. ; here is what I used for image. blonde female in pool with pouty lips with elegant body
    Negative prompt: bad-hands-5, badhandv4, NegfeetV2, easynegative,
    Steps: 28, Sampler: Euler a, CFG scale: 7, Seed: 3262718400, Face restoration: CodeFormer, Size: 512x768, Model hash: 4828b0bb4c, Model: jamJustAnotherMerge_v18Pruned (1), TI hashes: "bad-hands-5: aa7651be154c, badhandv4: 5e40d722fc3d, NegfeetV2: df90b1ff666d, easynegative: c74b4e810b03", Version: v1.5.1

    itsthehandAug 16, 2023

    Update: I fixed it I needed to change/add a VAE and it solved the issue for me.

    Purfecter
    Author
    Aug 16, 2023

    @itsthehand well yeah, that's why I wrote "There is no VAE baked in, so I suggest using either vae-ft-mse-840000-ema-pruned.vae.ckpt or kl-f8-anime2.ckpt with the merge." in the description. ^^

    1656015Aug 19, 2023
    CivitAI

    i am getting a lot of noise images too much color at the end result is really bad any help plz ?

    Purfecter
    Author
    Aug 19, 2023

    are you using a vae?

    1656015Aug 21, 2023· 1 reaction

    @Purfecter nope my bad it work fines now great job man love it

    Checkpoint
    SD 1.5

    Details

    Downloads
    471
    Platform
    CivitAI
    Platform Status
    Available
    Created
    7/25/2023
    Updated
    5/11/2026
    Deleted
    -

    Files

    jamJustAnotherMerge_v18Inpainting.safetensors

    Available On (1 platform)

    Same model published on other platforms. May have additional downloads or version variants.