Get Big Love Eryn1, Ultra5, Photo6, Zuna1 or Zia1 for $5.50 and Gwen1 or Hyper1 for $10 on Fanvue. Click 'Follow', buy a post, check your Fanvue DMs! You can get Klein2 for $10 on Tensor Art. A subscription includes the latest release and new ones within 30 days.
To create up to 85 free images per day in highest quality you can use Big Love here. For discussing photorealistic image generation join me on Discord.
Big Love Gwen is a finetune of Qwen Edit 2511 and can do NSFW image editing. It produces more realistic and sharp images with better skin texture & genitals than other Qwen checkpoints/loras. Txt2img is less reliable, but generates nice results too.
Big Love Gwen1 allows commercial usage without additional fees (if no auto-generation involved) unlike Big Love Klein. With its better anatomy it is easier to get good editing results. As fast at 6 steps as Klein at 4 steps.
The fp8 mixed version is for 16+ GB VRAM (highest quality), the nf4 version is for 12 and 16 GB VRAM (25% faster), the GGUF Q5_K_M version is for 16 GB VRAM (softest) and the nvfp4 version requires 24+ GB VRAM (on RTX 5090 2-3x faster, otherwise 2x slower, lower quality than fp8 mixed, less saturation, textured).
It is recommended to use ComfyUI. The Gwen1 workflows offer a skin enhance feature that is not available in Forge Neo. Move the Gwen1 model file into the diffusion_model sub folder of the models folder of ComfyUI, qwen_2.5_vl_7b_fp8_scaled.safetensors into the text_encoder sub folder and qwen_image_vae.safetensors into the vae sub folder.
For training a lora read here for Gwen1.
Settings are 6 steps (3 steps for upscaling), cfg1, Euler sampler, Beta scheduler, 832x1216 or 1024x1536 pixels.
Big Love Eryn is a fine-tune of Ernie-Image(-Turbo). You can run it locally with ComfyUI and Forge Neo. The fp8 version works with 8 & 12 GB VRAM. Use the (pruned) bf16 version if you have 16 GB VRAM or more. The "Full" versions are only meant for training a lora (coming soon!). Put the downloaded model into the diffusion models sub folder of ComfyUI or the StableDiffusion sub folder of Forge Neo. You additionally need to place ministral-3-3b.safetensors in the text_encoder sub folder and flux2-vae.safetensors in the VAE sub folder.
Settings are 4-6 steps (3-4 steps for upscaling), cfg1, Euler sampler, Beta scheduler, 832x1216 or 1024x1536 pixels.
Big Love Hyper supports generating at image sizes of 1664x2232 (4 megapixel) and higher with SDXL. Can be easily upscaled to 9 megapixel unlike the normal low-res images. Hyper currently has not fully stable anatomy, so needs more runs. Some prompts work better than others. It produces highly detailed and life-like images close to real photos. It has a special quality that cannot be achieved by just upscaling. The Hyper versions of Big Love are pay-only and special license conditions apply (see below).
Settings are DMD2 lora with strength 1.0, LCM Exponential, 8 steps, cfg 1, clip skip 2, 1664x2432 or 1536x2240. Upscale with img2img with DMD2 lora with strength 1.0, same prompt, LCM Exponential, 1.25-1.5x, 4-8 steps, cfg 1, 0.3-0.5 denoise, clip skip 2.
Big Love Klein is a fine-tune of Flux 2 klein. It can do txt2img but also edit images with prompts. Txt2img is currently experimental and requires more prompt & seed hunting. Image editing works stable. It only needs 4 steps, so is quite fast (5 sec on 5090, 15 sec on 5070). You can run it locally with ComfyUI and online on Tensor Art. Use the GGUF Version with 8 GB VRAM, the fp8 version with 12-16 GB VRAM and the (pruned) bf16 version with 24+ GB VRAM. The nvfp4 version (full nf4) renders 2-4x faster on RTx 5xxx GPUs with the latest CUDA drivers. The base version (full fp32 or full bf16) is meant for training a lora only. Put the downloaded model into the diffusion models sub folder of ComfyUI. You also need qwen_3_8b_fp8mixed.safetensors in the text_encoders sub folder (Qwen3-8B-Q5_K_M.gguf wth 8 GB VRAM) and flux2-vae.safetensors in the vae sub folder. Forge Neo is not recommended, because it currently produces too low image quality with Klein, which will hopefully change soon.
Settings are 4 steps (2 steps for upscaling), cfg1, Euler sampler, Beta scheduler, 832x1216, 1024x1536, 1280x1920 or 1536x2240 pixels
Big Love ZT, Zia or Zuna are fine-tunes of Z-Image(-Turbo) on the Big Love dataset. You can run it locally with ComfyUI and Forge Neo. The fp8 version works faster with 8 GB VRAM but produces more noise so may sometimes look more realistic, but details are not as good. Use the (pruned) bf16 version if you have 12 GB VRAM or more. The "Full" version is only meant for training a lora. Put the downloaded model into the diffusion models sub folder of ComfyUI or the StableDiffusion sub folder of Forge Neo. To make Big Love ZT/Zia/Zuna work you additionally need to place qwen3_4b.safetensors in the text_encoder sub folder and ae.saftendors in the VAE sub folder.
Settings are 8 steps, cfg1, Euler or DPM++ 2s a RF sampler, Normal or Simple or Beta scheduler, 832x1216, 1024x1536 or 1280x1920 pixels. For upscaling 8 steps, cfg1, Euler sampler, Normal or Simple or Beta scheduler, 1.5x upscale.
Big Love Ultra supports generating at image sizes of 1280x1870, 1360x1984 and higher. Can be easily upscaled to 6 megapixel unlike the normal low-res images. More photorealistic and detailed. A special quality that cannot be achieved by just upscaling. Ultra produces more details & sharpness when upscaling and detailing low-res images. The Ultra versions of Big Love are pay-only and special license conditions apply. See below.
Settings are DMD2 lora with strength 1.0, LCM Exponential, 8 steps, cfg 1, clip skip 2, 1280x1864 or 1360x1984. Upscale with img2img with DMD2 lora with strength 1.0, same prompt, LCM Exponential, 1.25-1.5x, 4-8 steps, cfg 1, lora strength 1.0, 0.3-0.5 denoise, clip skip 2.
Big Love Photo and Insta1 are finetuned versions of XL. Lust1 is a finetuned version of Lustify merged with Photo. They were trained on thousands of images and dozens of new concepts. Why they are the most realistic and versatile. For more details and prompting Photo1 read this article.
Settings are DMD2 lora with strength 1.0, LCM Exponential, 8 steps, cfg 1, clip skip 2, 832x1216 or 1024x1496. Upscale with img2img with DMD2 lora, same prompt, LCM Exponential, 1.25-1.5, 4-8 steps, cfg 1, lora strength 1.0, 0.3-0.5 denoise, clip skip 2.
Big Love XL is a combination of 5 different training branches of SDXL: original SDXL, bigASP, NatVis, Anteros and Pony. With a bit of Pony in it, some Pony loras (poses, characters) work too. It outputs a more photorealistic & creative look than the Pony versions and can also do paintings & cartoon.
Settings are DMD2 lora with strength 1.0, LCM Exponential, 8 steps, cfg 1-1.5, clip skip 2, 832x1216 or 1024x1496. Upscale with img2img with DMD2 lora, same prompt, LCM Exponential, 1.5-2x, 8-12 steps (XL1/XL2), 4-8 steps (XL2.5-XL4), cfg 1, lora strength 1.0, 0.3-0.5 denoise, clip skip 2.
Big Love Pony produces different images than both Pony & SDXL-based models. Pony tags are understood, but interpreted a bit differently. Some images lean more in the one or other direction depending on the prompt. It works with Pony loras as well as normal SDXL ones. It can turn anime/cartoon into photorealistic images in img2img. Pony3 includes a bit of Illustrious.
Settings are DMD2 lora with strength 1.0, LCM Exponential, 8 steps, cfg 1-1.5, clip skip 2, 832x1216 or 1024x1496. Upscale with img2img with DMD2 lora, same prompt, LCM Exponential, 1.5-2x, 8-12 steps, cfg 1, lora strength 1.0, 0.3-0.5 denoise, clip skip 2.
ComfyUI Workflow:
Download one of the following images and drag'n'drop it onto Comfy:
Klein, Eryn and Gwen workflows are embedded in the showcase images.
Z-Turbo/Z-Base txt2img & Upscale Workflow (recommended)
SDLX DMD2 txt2img & Upscale Workflow (recommended)
SDXL DMD2 txt2ing & Upscale for Big Love Ultra
SDXL DMD2 img2img Workflow
SDXL Lightning Workflow
Trained Concepts
Newer versions of Big Love include the concepts of older versions, e.g. Photo4 can do everything that Photo1 to Photo3 can do and more. The only exception are the Insta concepts, but they have been partially added to never versions too.
Photo6/Ultra5: ai-style, bodyscape, butt plug, candid amateur, cumshot, illustrious-style, outdoor sex, partner fingering, photoart, innie pussy, ray-style, rebellious, reflection, sensual, sexy ass, sweetheart, wedgie
Photo5/Ultra4/Hyper1: alluring, anal sex, artistic photo, artistic pinup, charming, cum portrait, cute asian, cute pinup, cute portrait, double oral, dreamy-style, face pov, fantasy pinup, fantasy-style, kai-style, lowkey, mj-style, outdoor flashing, photo art, pole dance, public indecency, public nudity, rock climbing, scify-style, sword pose, vaginal sex
Photo4/Photo4.5/Ultra2/Ultra3: 2girls kissing, acrobatic sex, average face, beautiful face, amateur pinup, anal fingering, artistic portrait, cuddling, cum swapping, doggystyle pov, even skin color, fashion photo, female masturbation, female pov, finger licking, food porn, footjob, french kiss, full nelson, funny movie, funny tongue out, insta cute, legs pov, low contrast, mirror portrait, nude pose, oral pov, orgasm face, outdoor nude, penis licking, pinup photo, pro photo, pro portrait, pussy licking, sex pov, sexy portrait, vaginal fingering, waterfall
Photo3/Lust1/Ultra1: 360 degree photo, 3d selfie, 69 position, adorable, alt beauty, anal gape, anal sex, average face, beach life, beautiful, beautiful face, candidness, cum, cute, deepthroat, dildo insertion, double exposure, enthusiastic, extreme pose, faces, fashion photo, female pov, fisting, follow me pov, girl next door, gorgeous, handjob, heroin chic, huge feet, huge hands, huge nipples, light rays, light streaks, lighting, low/mid/high contrast, lowkey/midkey/highkey, low/mid/high saturation, middle finger, motion blur, object insertion, detailed eyes, teeth, pro photo, pussy gape, real life, rooftop, round ass, sad, sexy pose, skin texture, skin tone, spread pussy, squirting, street portrait, testicle licking, testicle sucking, vaginal sex, voluminous hair, water splash
Insta1: amateur photo, fashion photo, insta selfie, pinup photo, pro photo, real life, big ass, bimbo, cute, luxurious, adorable, enthusiastic, average face, beautiful face, detailed eyes, sexy legs.
Photo2: 3d selfie, 69 position, adorable, alt beauty , amateur photo, ball licking, ball sucking, beach life, bubble butt, low/mid/high contrast, dildo insertion, double exposure, enthusiastic, extreme pose, female pov, follow me pov, forced perspective, girl next door, gorgeous, handjob, heroin chic, huge feet, huge hands, huge nipples, light rays, light streaks, lighting, middle finger, motion blur, no tan lines, object insertion, pussy gape, rooftop, sad, low/mid/high saturation, sexy pose, squirting, street portrait, water splash.
Photo1: amateur photo, pro photo, lighting, lowkey, midkey, highkey, beautiful, cute, candid, real life, average face, remarkable face, voluminous hair, skin tone, skin texture, 360 degree photo, spread pussy, deepthroat, fisting, anal gape, anal sex, vaginal sex.
Training a (character) lora
My advice is this: Take 10-30 high quality images (at least 50-100 for styles or poses), download OneTrainer, click the SDXL lora preset, choose Big Love as the base checkpoint, add your images as a concept, add tags on the Tools tab, and start training. If you train on Big Love Ultra change the training resolution to 1536 and provide 1248x1832 images.
You will only get the full Big Love quality by training on real photos or extremely photorealistic images, which is usually not the case with a fictional character. It is all about image quality and not choosing boring images. Don't add posing images to a character lora, because Big Love provides the posing later for it. Focus mainly on good portraits with a few upper and full body images. Rather fewer higher quality images than a lot of average images. Read here for traing a lora on Big Love ZT2, ZT3, Zia1, Zuna1 or Eryn1.
Fixing Anatomy
If you want to keep the seed and repair anatomical problems in an image, activate the Extra checkbox in A1111/Forge and set Variation Strength to 0.01 to 0.1. Then generate until anatomy is fine. In Comfy there are sampler nodes that support variation seed (e.g. Inspire Pack), which do the same thing. Also possible to change steps or add commas to the prompt, but I do not recommend it. Extra/Variation Seed is more powerful and convenient as it gives you endless variations with just one click. There is always a variation which looks similar or even better with great anatomy.
You don't need negative prompts for this. They are rather ineffective. You can also change the positive prompt to suppress things. Anatomy problems are also created by the prompt if you prompt contradicting poses that the model cannot combine. Better to fix the prompt first then.
Reducing Saturation, Warm Colors and Sharpness
These SDXL loras reduces the saturation and change the color temperature without reducing image quality. The lora strength defines the effect. Negative values increase saturation or add more warmth. To make an image softer use a lower strength for the DMD2 lora, e.g. 0.95, 0.9, 0.85.
Big Love Hyper/Ultra/Photo Sizes
Big Love Hyper supports these image sizes:
1664x2432 (recommend)
1536x2240
Big Love Ultra supports these image sizes:
1360x1984
1280x1872 (recommend)
Big Love Photo supports these image sizes:
1024x1496 (recommend)
832x1216
You can also use the 16:9 or 2:3 versions of these image sizes as well as landscape orientation, but the above should be more reliable.
The higher the image size, the lower the probability of good results. If you get too many problems with a higher image size, switch to the lower one. Some prompts do higher sizes fine, some do not. Do not use a lower image size than the recommended one, otherwise image quality will be lower. With higher image sizes anatomy problems occur too often.
Remove fill words and unusual tags in the prompt. They can cause a quality reduction with Big Love Ultra/Hyper as they were not trained. Shorter prompts should work better as it is less likely that bad or untrained tags are in it.
SFW Output - No Tits Please
"Is there a trick so that Big Love doesn't always generate raised shirts or visible breasts?" I get this asked so often, while it is so obvious. Big Love knows what men want so it gives it to them. Negatives don't work that well for it. So you got to speak to it in the positive prompt like you were taking to a nun. "Breasts? How dare you speak such an obscene word to me." Don't mention breasts or any of these sexy parts. Instead describe clothes a lot. Don't write down your impure thoughts. God forbid! Yeah, I know, so hard. Describe her body shape, which also implies certain tits underneath the clothes.
XL Prompting Advice
SDXL prompts & prompts from various SDXL checkpoints work with it. Pony prompts too, but they create a more photorealistic look. Natural language prompts tend to create artistic glamour images with less skin detail while normal tags do more photo-style images. Big Love XL gives individual words more attention, so you don't need to rework prompts like Big Love Pony requires sometimes.
XL1/XL2 has a tendancy towards cuteness and bright skin. If you get too much of it, remove words like cute, sweet, pale/fair skin & flash from the prompt. It also creates realistical analog photos with reduced image quality. Sometimes better to remove terms like analog or grain to increase quality. Decrease cfg to get a more natural look.
XL2.5 can sometimes get too sharp. Reduce cfg or use less steps in img2img to make images softer.
Pony Prompting Advice
Pony prompts & prompts from other checkpoints work with it. Score tags and fill words like masterpiece or perfect skin have as good as no effect usually. A negative prompt is often unnecessary unless you want to make something vanish or avoid underage, but it can also change the style of the image.
Pure booru tags and some other samplers may produce Pony style. Photographic terms or a simple prompts ensure a photo look. Sometimes it refuses to produce medium shots, portraits or close-ups unless you weigh the term heavily or remove feet, shoes, high heels, boots etc. from the prompt. wavy hair sometimes does not look so great.
If it refuses to do porn, try simplifying the prompt. If you want to do a portrait, don't describe surroundings too much. If you want to do porn, describe the action mainly and don't focus on the people or surroundings in the prompt. The largest part of the prompt should be about the main subject of the image.
If a prompt does not produce the intended image, you often only need to modify it a bit to make it work perfectly.
Using Character Loras
Big Love Pony supports Pony characters, but they look like real humans. There were some complaints about them not working, but the problem was not the model itself so far. So here are some tips:
* Try Pony as well as SDXL character loras
* Use a lora weight up to 1.5 if it does not reduce quality
* If it is a well known character, add the name in the prompt and also give it a higher weight if necessary.
* Check your negative prompt for problems. Delete it if necessary and rewrite it from scratch.
* Render a lot of images to be able to pick those with the best resemblance.
More SDXL Photorealism:
You can highly improve the images generated with this model by using img2img, upscaling, detailing etc. The SDXL Lightning lora enhances the look and constrast despite only 8 steps. Alternatively, the SDXL DMD2 lora produces a sharp, but more natural look with less contrast. Also check out the Subtle Style loras, which nicely enhance images.
Big Love License
Big Love SDXL versions are licensed under CreativeML Open RAIL++M (dated July 26, 2023), the Big Love Z-Image, Ernie & Qwen versions are licensed under Apache 2.0, and the Big Love Klein versions are licensed under the FLUX Non-Commercial License with the following additional terms:
For clarity, the following version tiers are defined:
Restricted versions: Ultra, Hyper, Klein, Zia, Zuna, Eryn, Gwen
Open versions: Pony, XL, Photo, Insta, Lust, ZT
1. You may use Big Love without crediting the creator.
2. You may sell the images that it generates.
3. You may only run the official Big Love versions of the SubtleShader account on Civitai and Tensor Art. Running it on other public or shared servers commercially (including other Civitai and Tensor Art accounts) requires a separate license or permission from the creator.
4. You may not sell this model or merges of this model.
5. You may not merge, share merges of, or share LoRA extractions incorporating Restricted versions. You may merge and share merges of Open versions.
6. When sharing the allowed merges of Open versions, as well as finetunes of them, you must apply these same additional license terms.
7. Redistribution of Restricted versions is prohibited. They may only be obtained from official sources designated by the creator.
8. All use-based restrictions of the CreativeML Open RAIL++M license apply to SDXL versions and their derivatives. All use-based restrictions of the FLUX Non-Commercial License apply to Klein versions and their derivatives.
9. Any restrictions in these additional terms may be waived only with explicit written permission from the creator.
TLDR: This is basically the same license as before (just stated more clearly), so nothing changes for the Pony, XL, Photo, Insta, Lust and ZT versions of Big Love. However, restrictions were added for Ultra, Hyper, Klein, Zia, Zuna, Eryn and Gwen versions (no merging, no redistribution). The "Share merges" condition below the license on the right hand side was only deactivated because of the restricted versions. Merging is still allowed for the open versions.
Many thanks to the creators of SDXL, Z, Flux klein, Pony, bigAsp and Lustify for making their wonderful models available. Big thanks to RaymondLuxuryYacht for allowing me to train on his fabulous images.
Description
More realistic, more texture, more natural skin color, more contrast, sharper details, faces a bit more mature, works better with DMD2, warmer colors
FAQ
Comments (158)
Absolute BEST, just real girls, not army of clones like other models.
V1+beautiness=v2.
I will try with celebs:)
WOW! My favorite model in the new version. I don't want to test other SDXL models as this is the best real masterpiece.
I think I posted this before, but any way to mention the size of a woman's breasts without them actually showing up? Any time I type medium/large breasts in the prompt, they're always exposed. Putting NSFW in negative prompt does nothing either.
Curvy = big breasts, slim = small breasts. Get creative! Mention a lot of clothes to avoid nudity. Add clothed.
I made a sound I am not comfortable with when I realized this was up...and I was just going to sleep...hahaha Thanks! Haven't tried xl2 yet, but your work is always next level!
if you are talking about realistic nudity , no model can do some certain body features , sorry for example butt cheeks , nothing realistic , and many more errors in female anatomy in all models , not speaking about yours only , hope you can take care of and fix later .
Only talking of photorealism. Flux will be better than SDXL at some point. Is not yet.
@SubtleShader 'Is not yet' ture.
FINNALY!!! futa update!! thank you!! I tried to make my own lora for tgirls on previous versions, but I couldn't make fine lora
Not intentionally 😆. Must have come with one of the new merged checkpoints or more Lustify/bigASP clip.
yeah;-) no need for the (great) TGirls Lora anymore. sometimes pussy appears but every third generation is anatomically correct
Without a doubt, the best realistic model for pony ever made! Like holy crap this is insane. This is like SDXL level of real but on pony now. Check out my profile to see what I mean!
yea this one really shocked me, even the prompt adherence is amazing, and this is my go to for ultimate realism, it beats every single xl checkpoint imo, especially with so many loras for pony out there to help enhance the realism, and if you're looking to work with complex prompts, this is the one
I know, you are not interested in it anymore, but you could still improve your Pony version. Just kick out the very overtrained Tame and use Cyber instead. It would be much more lora compatible.
Thanks. Will give it a try. I'm interested in making Big Love more photorealistic. Means Pony can only be a small part of it or the photorealisn decreases. I chose Tame because it is more photorealistic than most other Pony models.
@SubtleShader Thanx. It would be maybe less realistic, but worth for the sake of compability, adherence.
DRAWTHINGS (Tips for those using Drawthings)
You can still use the 8-step lighting lora (recommended). Drawthings doesn’t have the same schedulers as per the listed schedulers for this lora. Not to worry...
3 tested schedulers that work >>
- Euler A Trailing : softer image but good
- DDIM trailing : adds more detail
- DPM++ 2m Trailing : lots of detail but makes it up as it goes along especially with clothing
1) After generation(s). Switch to the ‘Scripts’ tab on the left hand panel.
2) Install the Single Detailer. (face improvement) if you have not already done so.
3) Run the detailer.
4) Upscale with either 2 or 4 Real Esrgan
_____________________________________
Big Love feedback
First of all thanks!
Personally, I think it’s the best pony checkpoint for realism and I’ve tried the main ones ie) Pony Realism, Cyber Realistic, Pony Dream etc. I’ve had issues with all of them, but that might be because I’m using Drawthings due to limited computing power. I’m in no way critising the model makers… I couldn’t do it. Big Love delivers a better result on my configuration.
I use same detail lora as the mod maker (detailed No trigger set at 100%) with the additional lora of Real Humans set at 120%. For Drawthings users, stay clear of cfg 3 or above. Anything 1 to 2.8 is ok. The higher the CFG the weirder the contrast. You go to 5 and you will generate weird red arms and burnt images… so don’t do it.
Negatives
As general with Pony. Prompt adherence. No matter how specific I write prompts (and I do write prompts like code > correctly layered, line breakts etc.) Clothing and complex prompts are a lotto when it comes to two or more characters. I expect that’s because Pony was based on anime and the tagging system was a generalisation of a visible image which already existed. The tagging system didn't list WHO was wearing what. When it comes to AI interpreting the tagging system it becomes the wild west. 2 in 8 or so images come out as visualised in my head.
Can’t wait for a FLUX/Pony merged model.
I'm skeptical about using Lightning with anything else than the SGM Uniform or Turbo scheduler. If DrawThings supports the LCM sampler, better use the DMD2 lora, which is more photorealistic than Lightning anyway.
You talk about Big Love Pony2, right? I'm only using Big Love XL2. More photorealistic & still has a bit of Pony.
Not possible to merge Flux with Pony. Incompatible. And it seems that Pony 7 won't be released for Flux. You can only train Flux on Pony images, which is not really a good idea.
DMD lora - to you have to link, i will check that out. Yes, Drawthings does support LCM. All the schedulers I listed do work though and have their own quirks. Yes, I’m using Big Love Pony2. I’ll check the XL2 version out. By ‘a bit of pony’ does that mean the same prompting or do you need a stack load of loras to make it do anything more interesting>?
@SubtleShader what at pity
@daveryder138 DMD2 link is above in the description. You usually don't need any lora with Big Love. Unless you want to do rather amateur or special artistic stuff.Too many loras & high lora strength reduces quality. Would rather keep number or loras and lora strength low.
@SubtleShader "trailing" in drawthings uses sgm uniform.
I still prefer Lightning 8 step to DMD2. I will definitely revisit and try some different settings but I feel like I couldn't find a sweet spot between too saturated and too amateur photo looking. I just posted two images with an identical workflow with Lightning but XL1 vs XL2, and I'm happy with them for different reasons. The really soft settings for 1 seemed to come out more balanced on 2.
@tsra Depends on what you want to achieve. Lightning has more contrast but is less photorealistic. Some images look very similar with both DMD2 & Lightning.
@SubtleShader Okay gotcha. By the way, it's obviously a very good NSFW model, but I also really appreciate how well it does certain styles and clothing (*cough* pasties) that aren't particularly obscure concepts but for some reason are not doable a lot of the time.
@tsra Big Love XL1 and even more XL2 combine a lot of different base models, which make it very flexible and capable of producing a lot of concepts.
For 2 or more characters you must use controlnet or regional prompter, if you want similar good details.
Did I read that Civit is actually going to enable on-site for all checkpoints this week as part of their Creator Payout revamp? Could you please make sure this one gets enabled for Pony if they do that? Thanks!
Yes, you are right. Thanks for the Info! Sure, will enable it if there will be an option.
Unfortunately no DMD2 on Civitai anymore. And they don't offer a LCM sampler for it anyway.
@SubtleShader so, what would you recommend as an alternative?
@aimediaartist Generate on your PC. If that is no option, TensorArt supports the LCM sampler. See my link in the description above.
This might be the only realistic pony model that can actually do wet hair. And it's really fucking good at it.
I can't get platinum blonde hair color, and I cannot get fair/pale skin tone, like pale British skin. Any tips?
Try white hair. Pale skin should work though. Check if there is nothing else in the prompt that might negate it.
Also 'latina' in negatives should help.
New Lustify is out, XL3 when? :D
New Lustify looks good when merged into Big Love. Improves composition and lighting. But I plan to wait at least till NatVis 3 is available before, which is announced for this month.
If you are an adventurer, you can try this clip
No, its not for XL or Pony, its for Flux, but works surprisingly well with some downside.
The realism is next level, but i tried with Cyberrealistic pony and couldnt make normal background.
Its like prompt adherence is on turbo mode. For example i use style loras with minimum strenght, but this turbos them to max.
I use realistic, realism etc prompt, and hell, its realistic as fuck. Skin texture on other level etc.
Good try.
Thanks! Will give it a try. You use it with Flux dev?
@SubtleShader No, i used it on Cyberrealistic Pony. It worked as i wrote, couldnt make a normal background.
It behaves, like a turbo adherence mode. It understands prompts in a literal way:) Can be too much.
@mussajosiah784 So you used it with SDXL? Just added it on the text encoder field in Forge? Or used it as clip in Comfy?
@SubtleShader Yeah, it works with XL and Pony too. I just loaded into text encoder in Forge. But there is a workflow for Comfy too
@mussajosiah784 Tested it. Produces different images that are similar in quality, but does not understand some concepts that the Big Love XL2 clip does and produces more standard poses. Makes no sense to replace the Big Love clip with it. Does not produce sharper results. Maybe improves Cyberrealistic Pony, but not Big Love.
@SubtleShader Does Biglove has custom clipL?
@2u0v645oz5683 Yes.
@SubtleShader Could you share it?
@2u0v645oz5683 it is baked into the checkpoint. Not sure if there is a node for ComfyUI that lets you save it as a seperate file.
Can someone explain to me how to turn an SDXL model into a video? Is this with Hunyuan?
Point me towards any workflows or articles would be super helpful
There are various image to video (i2v) models that are uncensored and run on a PC: Wan 2.1, Hunyuan, LTX, CogVideo etc. Wan 2.1 seems to be most advanced at the moment. Google it or search for it on Youtube. Wan2.1 is quite convenient to use on TensorArt.
@SubtleShader they all require comfy though right?, we can't do any good img2vid on Forge as such?
@makaveli_313 Yes, Comfy. Or do it on TensorArt.com online.
Hello
Thanks for this model. I have been trying to use your model with a character lora to good results.
My only issue is that it seems to largely work only with the resoulution 832x1216 or 1024x1536, as is given your description. If I go with a more 'traditional' 1024x1024 it starts to get issues with body parts or prompt adherence.
My question, would you consider offering 1024x1024, or perhaps even 1152x896 compatibility in the future?
Thanks for your work : )
Big Love works perfectly well with other image size. So I can only suspect that the lora is the problem. Or it is not compatible with Big Love at other image sizes. Or maybe you just need to modify the prompt.
@SubtleShader I used this lora with other nsfw models(d33PSIX, monolith6) and it seemed to be compatible with said resolutions, but as you mention it I will have a look again.
Could you instead tell me how to have a 'zoomed out' image render?, I've tried prompts such as long shot, medium shot, shot from a distance, medium crop etc. etc., but none of them seem to work and the images are all super 'focused' in on the subjects. Where the image output always has end to end the body of the subject(s) in image.
If you can suggest prompts or perhaps lora's that might help with getting a zoomed out output where there is more of the surrounding environment in the shot It would be helpful.
Thanks again.
@makaveli_313 Try full body
@SubtleShader oh perhaps i did not explain well. the body is already fully in frame, i'm trying to make it so the output image isn't so cropped into the body.
i.e. the body as is fills image from top to bottom, im trying to 'zoom out' more. much more, if possible.
@makaveli_313 Never tried that. Bodies don't get rendered well then.
@SubtleShader alright thank you.
Other than that the model works very well with my lora. Looking forward to your future xl updates if any : )
@makaveli_313 XL3 in the works...
@SubtleShader looking forward to it!
although the model(atleast with my lora/usage) is so damn near perfect, what would be left for you to add?
@makaveli_313 Checkpoints of the Big Love merge being updated... merge more or my loras etc.
@SubtleShader and when you say Merge it means more out of the box?, i.e. no need for detail lora's etc. etc.,
and perhaps more character face variety i presume?
@makaveli_313 For Big Love XL2 combined 8 checkpoints that are very different to add the best of all of them. And 4 loras that I trained myself. 5 of the merged checkpoints were trained from scratch, so there is a lot of variety in Big Love. Detail loras are not necessary. They often make it worse or have no effect at all. Check out my own loras, which let you add a totally different style.
@SubtleShader Thanks for the feedback and help subtleshader!.
Awesome. Just awesome. The colors have improved so much! I've used this as a base to train a LoRA with pictures of my girlfriend. The results are fantastic.
One small issue: every now and then, a third leg appears in v2
Nice to hear. These anatomical problems are normal for SDXL. The more people in an image, the worse.
So happy with this checkpoint that I released my first LoRA based on it and pictures of my girlfriend:
@fqxdcptgsuqdinipxu718 Congrats on the girlfriend. I hope she was Ok with it. Personally I would not release such a lora, because some dudes could produce crazy images with the lora.
@fqxdcptgsuqdinipxu718 Are you sure that the Civitai Terms of Service even allow posting a lora of a real person?
@SubtleShader You're right. Although she doesn't mind, it's not allowed by the terms of Civit. I'll delete it. Thank you for pointing it out
Praise the sky gods! Big Love is available on Civit on-site generator! 🥳🎉❤️🍆💦🍑
Unfortunately quality is not great on Civitai. Better locally or on TensorArt. Hopefully will improve in future.
@SubtleShader is there any reason why? I'd love to get good quality generations on civitai! I'll try tensor.art
@PixelPerv Civitai does not support the LCM sampler for DMD2. Also not the SDE sampler. But even then It looks worse than what you get with A1111/Forge and Comfy when using the same settings.
@SubtleShader Sadly, I don't have a Windows machine. I'm also worried about using Tensor art, people say it is sketchy.
@ponyboi TensorArt seems to produce more failures because of fp8, but ok. Nothing beats running Forge or Comfy directly. Can also be done in the cloud, but more expensive.
@SubtleShader Even without living up to it's full potential, running on Civit has been fun so far. I like the challenge of pushing base models beyond their capabilities with interesting LoRAs combos. Doing that within the confines of Civit increases the difficulty which keeps it interesting. Thanks for making it!
Thanks @SubtleShader, it's given me a reason to try out Forge on cloud GPU, the results are great! Big Love was recommended by the creator of the Real Cum lora.
...and it's gone.
@ponyboi Popular move
Someone asked me how to prompt certain sex stuff. Here is what I found works:
1girl, 1boy sucks nipple, breastfeeding
1girl, 1boy, handjob, holding penis
2girls, 1boy, threesome, sex
1girl, anilingus, tongue in anus
Unfortunately very hard to combine 2 of these. Would require training a lora.
Alternative use a Pony based model in txt2img and Big Love in img2img.
For examples see https://civitai.com/posts/14111266
SD 1.5 Loras don't work with this right?
Nope. As it is a SDXL checkpoint, only SDXL loras work with it. Also Pony loras, but they are SDXL too.
Had a look at Flux again after I stopped using it mid Oct. Still very tedious to use, poses & perspectives mostly boring, ignores NSFW prompting, often waxy look... all in all, nowhere near as flexible, creative, photorealistic & fun as SDXL. And Flux renders 4-6x slower.
Makes no sense to do a Flux version of Big Love. I don't want to bore anyone.
Maybe they will manage to remove the censorship in the T5 clip & make it worthwhile. Otherwise some other model will likely take the lead in future as far as NSFW is concerned.
yeah totally. I agree. sdxl is the future. I actually made a merge between BL2 and OLT and I'm very happy with the results. I'm sure BL3 will be groundbreaking. Cheers.
@ElectricDreams Small steps forward. Already very pleasing what is possible.
agreed, flux and sd3 out and sdxl still the best. go figure
@kunde2 We would not be here without the guys behind bigASP & Lustify.
Did you see epiCPhotoFUN? Could be a very good addition to Big Love.
@ikwerner1432 Thought about merging in one of epinikion's models, but some don't allow merging and all are generally not NSFW enough. Don't want to reduce the NSFW capabilities of Big Love.
People are severely underestimating high quality SDXL models like Big Love paired with Wan I2V; for consistency and uncensored scenes, I really do believe that is the way going forward.
@ablestable69420 At some point we will need something even better than SDXL that is not censored. But I don't see any alternative yet.
@SubtleShader i don't think that is gonna happen anytime soon. Lumina -which was better than sd 3.5- was an uncensored model but failed to gain the attention of the community which is not gonna make an effort to train a new model unless is noticeably better than flux right from the scratch. Nobody is willing to spend 1,5 years in a model to git gud anymore (like happened with sd 1.5 or sdxl) . Not only you need a good model to be released (better than flux dev), the harder part is having the community invested in.
Flux was ruined from the start.. and having all those copypaste merges ain't helping its case, people think adding their training or (none at all) only obfuscates things beyond repair
@axicec Yep, the Pony developer said recently that the problem is not T5 when talking about Pony 7. So the real problem might be the initial Flux training. It likely has to be completely retrained from scratch. Same as was done with bigASP, which resulted in very creative models like Lustify & Big Love. But not that easy and much more costly with Flux. So Flux might be a dead end.
@ElectricDreams Lumina is censored. Blurred nipples. God never created women with blurry nipples. Poor babies & husbands. Should be forbidden by law to disdain God's creation. Sacrilege!
@axicec so true. It's obviously a poisoned model which was really obvious from the start so It never interested me. There's no way to fix that no matter how much you train it (your trainings are gonna collapse anyway). So train flux is an excercise in futility.
The answer is Pony7 or some new not anime base model from China with full of celebs (which Civitai wont host, but who cares)
@charlestewani114 Not sure about Pony 7. Slower than Flux at the moment according to the developer and photorealism worse than Flux. Might be a flop. A open source image model from China would be welcome. Fast, capable, uncensored und runs on consumer hardware.
@SubtleShader Hell, Pony7 hardware req is worrisome, but its capable for at least XL realism out of box. With finetunes more better. If people jumped on Flux, they will jump on v7 too. A big uncensored model with t5 like is a dream for tech people.
@SubtleShader lumina -afaik and in theory- is not censored, it's just not trained on nipples or nudity. Sdxl couldn't do nipples either at release. I don't think it's poisoned like flux. It could but at this moment we don't know because nobody is training on it. But the question is: it's lumina good enough to waste a year in finetunes? probably not. Something better -most likely- will be released on that frametime.
@ElectricDreams Should also not be too slow on consumer hardware. Any article about the Flux poisoning?
@SubtleShader it's well known that flux finetunes training collapses beyond a number of steps i don't remember know. But i'm sure there's information on reddit.
@SubtleShader So there is this uncensored model based on Flux-Schnell doing the rounds:
lodestones/Chroma · Hugging Face
@Melty1989 Have not seen any NSFW images from it. There is another one called Flex.
@SubtleShader I was able to generate NSFW with Chroma. However, that's mostly in terms of nudity and genitals. It can't do positions at all.
Didn't read all the comments... Still...
I know my pal @SubtleShader has no time to lose with images renderings. (his Comfy workflow is great when you are in the hurry for some AI Porn btw)
Everybody 's not expecting the same things from AI generation, some want extreme close-ups on ultra raw realistic cunts, some others want more scenic/crazy/creative and artistic "setups"...
and some want both. I want both, and like it or not, Flux seems to be the future... Our gear is just not ready...
Waiting for a better gear and better Flux CP, i'll keep doing flux txt2img then BigLove img2img to get these scenic/nsfw visions out of my head. ;-)
@BretChampagne I tested Flux & noticed that it is not on the level of Big Love & similar SDXL checkpoints. Had a similar experience with SDXL in Dec 2023. As you might guess I'm very demanding. However, there are some bigger stepping stone with Flux compared to SDXL. Another problem might be that Flux might never get fast enough because BFL might keep optimisation for their commercial pro version.
@SubtleShader Flux is way more capable than sdxl in many areas, and way more respectful of my overly twisted prompts...
I totally understand that you don't want to bother about a Flux version of BigLove, but I have to disagree when you write that Flux is less creative than SDXL. It's up to you to be creative.
@SubtleShader
For the same prompt...
SDXL > https://civitai.com/images/63957267
FLUX > https://civitai.com/images/63949410
You may like better the SDXL version, but the Flux one was exactly I had in mind when prompting ...
@SubtleShader I give you the Flux poses are boring as fuck ...
It does not even seem to understand "squatting" XD
@BretChampagne Less creative in terms of my twisted poses & perspectives. I see that Flux works for your style, but not for mine. Another problem for me is speed. I can be less productive with Flux.
@SubtleShader Yeah, I know what you mean. It takes me more than 3mn for one generation. Then 2 more for my overkill img2img workflow. Fucking 8Go Vram...
@BretChampagne 12 GB is not much better. Despite Turbo lora.
@SubtleShader And it takes me 45mn to generate locally 3sc of NSFW Wan video XD... But that's antother story...
I have never liked the skin tones on flux, waxy as you put it.
How is the next SD(3.5?) for NSFW?
@makaveli_313 No idea.
@SubtleShader I totally agree with your assessment regarding Flux. Thanks for keeping SDXL alive. It is a pity that Astralite of @PurpleSmartAI hasn't done a v6.9 on the SDXL architecture.
I would just like to mention the Pixart Sigma architecture here, which is heavily underrated in my opinion. It could have some advantages over SDXL due to DiT instead of the Unet if trained right.
Hi! What means, when a checkpoint only works with Euler a or some slow sampler(SDE, etc).
Overtrained?
I dont talk about your model, just in general.
There is here a vibrant life, thatswhy im asking it here:)
Does not mean anything. You should always check which sampler works best after creating a new model. SDE is slow because it does double steps, so might need less steps than others. Euler A works best with some Pony models, but I also used it in the past for upscaling with Lustify 2.
Sometimes overtrained is what you want. For example, with loras it might be good to overtrain & later use a lower strength. Can improve image quality.
There are no real standards. Just use what works best for you & your intention.
@SubtleShader I mean, i always use 2m, and dont like euler a. I speak about realistic checkpoints, which give heavy artifacts with 2m, 2msde. Why this happens, when most models works good with 2m?
@charlieyamasaki734 When merging models with others or adding new training, the weights change and another sampler might work better. Just do a xy plot after downloading a new model and see which sampler & scheduler works best. Usually the model creator already mentions it in the description or uses them in his showcase images.
How do I make consistent characters? Can I make a lora on FluxGym for this model? if yes how do i do it?
I use OneTrainer. Very easy to use. Fluygymn only supports Flux. Bug Love is SDXL. The OneTrainer defaults are quite good. Train on 10-20 images.
U have loras to recommend for most realism? Do u have any site/video to see how to use onetrainer, im pretty new to this? its ok for 12gb ram?
@TropicalCat Search Youtube for OneTrainer. I only use 12 GB VRAM myself. Check out my loras at https://civitai.com/models/1115264/subtle-styles
@SubtleShader Do you recommend any workflow in particular?
@TropicalCat What do you mean with workflow? ComfyUI workflow?
@SubtleShader Yes a comfyUI workflow haha
@TropicalCat I mainly use Forge. There is a ComfyUI workflow linked in the description that does what I do with Forge.
Getting great results, thank you!
Lustify End Game was unmatched for a while there Big Love now Big Love V2 takes its to all new levels, Best checkpoint ever made, the SDXL version being far superior, loads of scenes, styles, characters, niche subjects are missing in the Pony version. Thank you so much for bringing this to community!
This one takes the cake,
and how do I keep it from generating penises on females at random😭
Thanks for the buzz! Accidents happen.You either just sort out those images (you will get used to it!) or don't add penis or sex to the prompt. Or only prompt for lesbian sex.
Adding "futa, hermaphrodite, female penis," to the negative prompt has worked for me. Also, removing "solo" from positive has proved useful because I think the embeddings match "solo" and "penis" in positive with ladyboys and twinks.
Have you tried a gender slider lora?
Big Love XL2.5 available!
What's new? Darker, sharper, more contrast, more skin texture, fine grain. Higher DMD2 cfg without overexposure. Requires less img2img steps. See Comparison.
Is Pony getting anymore Big Love?
@kellykelly There is Pony in XL2.5 😁. The old Pony1/2 versions are less photorealistic, because there is much more Pony in them. I might do a Pony3 if a more realistic Poney model than TamePony 2.5 becomes available.
Is this good with img2img and inpainting? I tried Lustify for it but except v3 (which is removed by the author for some reason) all others are failure. I didn't try the latest lustify version though.
All versions of Lustify are good. No failures at all. You must be doing something wrong. Or you expect something that is not cannot do. I love Big Love. But no idea what you expect. Often I don't need to do img2img, because txt2img is already good enough.
@SubtleShader I probably doing something wrong then? idk what that is but ok.
@chrisss1 Try Big Love and let me know if you have any problems.
@SubtleShader Will do, thanks. Waiting a new version :)
You know that Inpaiting requires an inpanting model right?
@Dlzeze No, it does not. Works with the regular model too.
@chrisss1 try Lustify OLT without the texture fix - the original OLT. You will find success with both i2i and inpainting there, most likely. also, @Dlzeze , no, an inpainting model is not "required" for inpainting, where did you get that idea? In fact, in my case, since I was already a skilled inpainter before inpainting models became a thing, I can't get used to the wildly different inpainting settings required by inpainting models, and I just continue to use normal models to great success. It's all about the prompt and the denoising and the mask blur, but most importantly, the context window (often called "mask padding" or specifically in the case of the ADetailer extension for A1111/Forge, "Inpaint only masked padding, pixels").
@shapeshifter83 Thank you. As for inpainting, i have some questions for you if you don't mind. Since you been using inpainting and i'm assuming you are pretty good at it, can you recommend me some tips on how to seamlessly inpaint a character, object anything else in the original picture? Sometimes i get good results but most of the time the inpainted character doesn't blend good in the original image. I tried soft inpainting but it was a waste of time for me. I lower denoising strength but this time it doesn't change the character much etc. Thanks in advance
coyotte actually released an inpainting version of his OLT model
@kunde2 i'm aware. I don't use it, I still inpaint with the original OLT. The inpainting model requires setting changes that I'm just not adept with yet.
@chrisss1 try increasing "inpaint masked padding" by A LOT. That's the context window of the inpainting pass. It's what the painter "sees" while it's reading the prompt and painting at the same time. That's the best first step to managing blending, in my opinion. And then, you want to match your prompt to whatever is within that "inpaint masked padding" range (if you're using A1111/Forge, the number is in pixels). So if you're trying to add a person alongside an item and your padding now includes that item within it's range, go ahead and have the item described in the latter part of your prompt as well even if it's not what's being painted. This way, the painter sees the item in the context window and matches it up to the item in the prompt, and "sees" the picture more completely, which will help with believable blending. The gist of what i'm saying is ***match your prompt to your context window (masked padding) rather than the actual inpaint zone*** for best results. This will also allow you to create larger inpaint zones which tends to help with blending. Big inpaints with big context and a matching prompt tend to blend the best - the downside is that you'll often need to increase denoising to get a strong change in the image, which can often alter color saturation levels. There's always trade-offs, you just have to balance them. But yea, increasing the masked padding context window by leaps and bounds is really how I started getting good inpaints.
Another setting that can help with blending is mask blur, but only for the type of blend failure where you can see the outline of your inpainting zone. Otherwise you want to try to keep this as low as you can because it causes actual blurs or "ghosting" around objects or people.
Then there is the A1111/Forge "soft inpainting". I use it about half the time, and I generally find it to be pretty solid, especially combined with an increased context window. the default settings are pretty decent but sometimes i'll raise Difference contrast because it seems a little on the "ghosting" side and sometimes raise Mask Influence if it's getting a little too liberal with changing areas outside my mask.
Most of my inpaints are done with "fill" (where it samples from average image colors as a latent to inpaint on) and some are done with "original content". My denoising is usually 0.65-0.85 for the most part.
Can't think of any other tips or advice at the moment. Hope some of that helps.
@shapeshifter83 Thank you i really appreciate it, i will try those. Last thing, what sampling method do you recommend, i am using DDIM at the moment. And which one is best for realistic?
@chrisss1 i tend to go for quality over speed, so I use "Euler a" with a high step count - sometimes as high as 96 steps when I really want good stuff. Generally when I'm just searching for seeds in t2i i'm running with 48 steps. I have a pretty good video card though and I know not everyone has that luxury. I have used DDIM a little but it hasn't really suited me. There really isn't an identifiable "best" sampler, it really depends on your own preferences and your own desired balance of speed vs quality and the subtle visual differences.
@chrisss1 oh and one more thing on soft inpainting. Soft inpainting is a weird thing and I'm still not a complete 100% on how it's working under the hood, but I think I have a pretty good idea. It seems that every step, it's creating a brand new mask - more-or-less based on the mask you drew - with a variability and extension beyond the edges of the mask you drew, and this variability is partially random and partially guided by the various soft inpainting settings. Then it paints this new mask it created for that step, rather than the mask you actually drew. What this means, I think, is that the mask padding context window is changing every step and is much larger than the actual mask you drew. So for the most part, unless i'm doing a very small inpaint on a very upscaled image, I generally leave my prompt to include the entire image when i'm doing a soft inpaint. Like I said, i'm not 100% that's what it's doing, I'm just basing this off of trial and error and getting a feel for it's behavior. But I have certainly settled on assuming a full-image context window when soft inpainting, and it seems to be working out.
@shapeshifter83 Thank you, so Euler a then? i know and i heard that it is fast as well but not sure about realism. I will give it a try. My preference is balanced. I don't have a really good pc so i am trying to get between quality and the speed. As for soft inpainting, i didn't had luck with it, whatever i do it seems like doesn't work well.. at least not with DDIM i think. I will try it with Euler a. Mask blur helped a little bit and masked padding i think the most useful one.