Big Love Klein3 is also offered for $10 on Tensor Art. You can 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! A subscription includes the latest release and new ones within 30 days. Any problem? Contact me via chat.
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 Klein is a fine-tune of Flux 2 klein. It can do txt2img and also edit images with prompts. It only needs 4 steps, so is quite fast. Check the "About this version" box on the right for info on the individual variants. Some are up to 3x faster.
It is recommended to use it with ComfyUI. 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.
Settings are 4 steps (2 steps for upscaling), cfg1, Euler sampler, Beta scheduler, 832x1216, 1024x1536, 1280x1920 or 1536x2240 pixels
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. Check the "About this version" box on the right for info on the individual variants.
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 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, 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, Qwen, 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
Photo1 is more photorealistic, contrast & skin details were improved, the default face is prettier and freckles look better. Image quality & skin texture much better than using XL4 with a lora for similar concepts. 22 new main concepts were trained with 2150 images as discussed in this article , e.g. 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.
FAQ
Comments (170)
Hi!
What is the best setting for Hires fix, if i use dmd lora? Hires sampler, scheduler, cfg?
Best to use img2img as it produces higher quality by allowing more steps. With hires fix you need to reduce the hires steps. Otherwise same settings. Img2img also had a seed, so you can redo it till it looks best.
SubtleShader I use the settings, what you wrote. With hires the upscaled pic is simple very bad(especially eyes), with img2img i tried SD upscale, which is not that bad, but not good too.
It seems, the upscaler model matters a lot. With your upscaler model the eyes are too clean and not even close to the original pic.
Hah, thanx anyway. I have just pirated the new gigapixel, and even their standard model is awesome(2 sec, 2x upscaling)
rettodysseus382 I don't use any upscaler. Just plain img2img. Apart from showcase images, I usually don't upscale. Sometimes I detail face, breasts & pussy.
SubtleShader Thatswhy i didnt understand the img2img description:) Thanx, very good now with dmd.
Great model. Thanks for your work. I'd like to know if it's allowed to integrate this model in a merge along with 2 or 3 other checkpoints? I guess some models are copyrighted. I haven't found how to verify this information.
Click on the three icons below my user name near the top of this page and you will see. On some model pages there are no Icons and only a license displayed, which means that all options are allowed if not limited by the license conditions.
It's tricky to get rid of tan lines. Try "no tan lines" in negative or "no tan lines on [body part]" in positive. I've also had some success with adding "even skin color" to the prompt.
try (tanlines, tan lines, tan-lines) in neg. that helps me
This is the only model, where dmd not sucks, it gives great details, while other models suffer with it.
V5 when?
Thanks. The new Photo1 will be even better. Skin texture is great & the new concepts quite handy. Various stuff that was only possible with a lora does not require one anymore, which also means that image quality is better. I'm preparing the showcase images now. Should be out in a day or two.
SubtleShader Great!
Big Love Photo1 Released
Read the full story in my article here.
I only set the early access fee so high, because I rather prefer people to use TA download/activation. I get the same amount from the buzz as from the TA fee, but buzz is much harder to convert.
What is TA activation?
Great work man! Don't stop!
Dynelements TensorArt
Dynelements You can download it there too. Or create images with it there online. Here is the link: Big Love on TA
Thanks for sharing the gory details behind the training and work you've done. I just purchased early access on TA to help support. Very impressive what you've done with just a 4070! Can't wait to take Photo1 for a test drive.
shanonathon Thanks. It was more a time investment than investment in GPU/electricity costs.
@SubtleShader Great work. You gain me back again :)
The return of the King.
👑
I downloaded and tested all three models of BigLove Pony. There doesn't seem to be much difference in quality between the three models. However, models 1 and 2 display strange watermarks and logos. Model 3 doesn't have these issues. Therefore, I think model 3 is the right choice.
You are the very first person I ever bought a checkpoint from and I don't regret it. You deserve all the financial help considering that you put your own time (and money!) into this for free for everyone!
Thanks a lot. I'm too passionate about it. 😁
love the models. but question, how can i prompt for big breast but sfw, not showing them?
Describe a lot of clothes. Try curvy or deep cleavage instead.
@SubtleShader ye i think its like givin less power to the boobs idea right? and more to clothes or other things
@haruu_ Yep
Any tips on how to force these triggers: fully swallowed, half swallowed, chipmunking, gagging?
I made my prompt as barebones as possible. I had no extra Loras at all in my prompt except the Dmd2.
Added weights to those loras (1.2-1.5) and yet the images didn't seem to have noticeable resemblance to those triggers. I tested various dimensions as well but no luck (1024x1496, 1024x1024, 832x1216).
I am also struggling to get bridge position, piledriver position to work as well.
Pronebone works well though.
Bridge & piledriver position are unusual, so they would have required more training or training images than the other position. Will try to do that in Photo2. Otherwise it takes a lot of rendering to reproduce them. Upside down and button up top down might help with prompting piledriver.
Don't use loras with such high weights. Will produce bad quality. Try to avoid loras with Photo1.
fully swallowed, half swallowed, chipmunking, gagging only nudge probability in your favor, but they don't work every time. All checkpoints that were merged into Big Love are headstrong when it comes to Interpreting prompts. The upside is a lot of creativity.
How does this model deal with feet?🫣
I did not specifically train feet
My SDXL foot LoRA tear it the eff up though on this.
This is one of the best checkpoints I've tried, and it produces something very close to real-life amateurish quality. Outstanding work!
Man your models are amazing, i only have 8gb of vram and xl4 and pony3 with dmd2 were a game changer. The quality is amazing and they have a great lora compatibility. Pony3 is the best pony model for realism i was literally amazed when i saw the first results. I cant wait to try photo1
Thanks, man!
The eyes ruin everything in this otherwise awesome model. If i must use adetailer, the concept of fast generating is into trashbin.
SUBTLE?? - WTF Happened?!?! I want to claw living fk out of my eyes so so bad right now! ... :D
Before I start please accept this as purely constructive feedback from THE worlds foremost expert on Big Love Checkpoints! - Hands down - I know absolutely every inch of capability of Big Loves, far more even than you -
YOU who is the undisputed checkpoint fkn maaaaster!
so hear me out here! ... :D
We have been exclusively using Big Love at our agency since it's first release.
it was and still is the perfect checkpoint for what we do, there is nothing else that can touch it, until very recently Flux or any olther model / checkpoint out there could come close to the photo realsim of Big Love.
As such, at our agency we have generated over 62,000+ images on our RTX Pro 6000, 96GB VRAM workstation, solely using your incredible creation, as i say due to its unmatched photo realistic capabilities.
This is really of no consequence to what we do as we have our fine tune, with the absolute sdxl GOAT still to this day the untouchable - Big Love XL 2 - which features and additional 150,000 photo dataset which we would be happy to share with you.
This leads me on to the feedback of XL 4 - I have been away and returned to see the release of 4 - As you can imagine super buzzed to see a a new BL release - as you may or may not remember - I have said many times in comments on your work, that YOU are the checkpoint FKN maaaaster!
So I am hoping you will take this constructively! :D But in the name of all that is holy did you do to make this checkpoint series go from spitting out BY FARRR the hottest most photo realistic eye candy in the history of AI image generation , to the fugliest most putrid demonic , chromozone lacking dispicable beasts, dipped in intenslty over saturaed vomit inducing colorrs, wraped in AI plastic skin, with a very consistent "Flux face" esque sheen to them.
And we are talking here identical prompts, comfyui environment as was the framework for 60,000 images that came before it, we simply swapped out the models.
Well I suspect is that by the time version 4 came around it was it had been stuffed and jam packed so tight with checkpoint merges, that it has turned these b%"$es into HORRIFIC retched spawns of satan!
You are still and always will be the checkpoint fkn maaaaster! - i just wanted to feedback and share with you, the worlds most foremost Big Love experts - honest - opinion!
I am about to purchase the new Photo version. We will be setting up our models on TA this week, all our sdxl models have a big love base and expand on the the photo realistic aspects.
BIG LOVE to you subtle you absolute FKN Legend of the space!
Interesting. Personally, I thought XL4 was WAY better than previous versions of BL, especially in terms of textures and backgrounds. Maybe you just need to patiently tweak your prompts to achieve the aesthetic you are after (desaturated, pale skin, cute face ...?), try reducing steps or CFG (unless it is already 1.0) or maybe just keep using the checkpoints you already love? I have yet to try Photo1. Thought 2: I think XL4 is optimized for the dmd2 speed LoRA - are you using it?
It seems, you dont use dmd2 with version 4. Its a must have, tones down the saturation, colors and contrast. Its clearly not usable without dmd, despite, its not baked in.
Thanks for your feedback! XL2.5, XL3 and XL4 are basically XL2 plus some loras and weight manipulation. I'm trying to optimize it for the DMD2 lora this way. I did not find any checkpoint to merge since XL2 that would improve it. Photo1 is a trained and merged version of XL4 thus has the same stronger contrast, saturation & sharpness. I understand if some people don't like that and prefer XL2, but most moved on and adapted newer versions of Big Love.
I could create a version of Photo1 that is a merge with XL2 instead of XL4. Will have a look at that. Do you use Big Love XL2 pure or use it with Lightning or DMD2? I'm using DMD2 since XL2.
Yep, would be great to have access to your 150,000 photo dataset. Would certainly be handy to select some of them for training future versions of Big Love Photo. Open a chat or contact me on Discord for that. Thanks for your fun comment!
@shavarzorawar432 Yes, XL2.5 to Photo1 are optimized for DMD2. But can also be used with the Lightning lora, which I sometimes do for inpainting & img2img. Did not check it without both loras much, but noticed that some prompts don't work very well without. So best to always use DMD2 with it.
This is very scary.
Calling bullshit on genning that much and suddenly not being able to figure out XL4. I've gen'd that much (honestly likely much more) on Big Love and have not experienced a single spawn of satan with XL4. Sort of moot since Photo1 is a massive jump overall at what XL4 did, but I highly recommend you check your workflow to make sure you didn't shift a 1 or a 0 somewhere without realizing it. It happens to the best of us.
@DrAIvyn He is probably not using DMD2.
@shavarzorawar432 Dmd2 is just a lora? Is there anything else it means
@JetterJack dmd2 is a LORA that when applied with the right settings (Sampler: LCM, Scheduler Exponential, Steps: 6-12, CFG: 1.0-1.4) enables you to make SDXL images really fast. If you have not tried you really should. Much quicker. Download and apply the LORA, change the settings.
@JetterJack DMD2 does not just speed up calculations by 3-6x. It also produces a more photorealistic look.
I was having trouble getting quality results with Photo1. After getting a handle on the new features(lighting, realism, etc) and retraining my character LoRA, it's quite good. The only question I have with Photo1 is about the Subtle LoRAs. Specifically the "Subtle Poses" one. Are the concepts from that already baked into Photo1? Regardless, great work. Really enjoying Photo1.
Yes, 50% of the training images of the Subtle Poses loras were trained on Photo1. But also from various of my other loras. So you normally don't need these loras anymore. Avoids image quality reduction by loras.
@SubtleShader "Avoids image quality reduction by loras."
This is exactly the reason I was asking. Thanks for the confirmation!
Use PhotoFlow XL DMD v2.0 workflow for best results. This checkpoint is good!
Is there any special in the workflow, beside facedetailer and upscaling?
Most of these workflows are easy to copy in A1111, Forge too, but nobody wants to explain, what to use.
Thanks. Will check it out. Also look at 2 other workflows. Then probably come up with an improved one for Big Love.
@jashunmoritz947 Just look at them in Comfy, see what they do and then try to replicate in Forge manually.
@SubtleShader :) I wont install that piece of software, only when pony7 doesnt get support in Forge.
@jashunmoritz947 It won't
@SubtleShader Ill need a whole new machine for it anyway. They are not in hurry, fully uncensored models will be very rare.
I was trying to buy your thing on Tensor art, but I got a 404 error.
Same here! :(
I contacted them. Should be fixed with 24h. Happened once in the past. No idea why. Sorry.
Big Love is visible again on Tensor Art. Sorry for the incovenience.
I’m trying to use InstantID to extract a face, but I get absolutely no resemblance regardless of the parameters. Does anyone have any advice for BigLove XL4 with the DMD2 LoRA?
use pulid with 0.7 str fidelity and 10-12 steps
@sw888 Do you know where I can get a workflow to mess around with this?
@anonameguyman1234252 you should go to pulid for comfyui github page and follow their instructions. for the workflow, simply spawn the apply pulid node and connect what's missing. but you can find some online. i'd send you one of mine but i have no idea how
Why in earth I got this messy result? https://iili.io/KKzAdRS.png
Looks like when I used 25 steps with DMD2 instead of just 8.
More like normal cfg 5 something.
My only problem with low steps, low cfg, that clothes are unfinished and different materials have no time to shine.
You managed to perfect the skin with dmd, would be hard to put more emphasis to materials and clothes?
Sorry, never tried. Might add some clothes training sets in a future version. You could try using more steps with maybe reduced DMD2 strength.
will this model run on 4gb vram?
@ash6999 Only if you use the fp8 mode in Forge or use certain nodes in Comfy. Only tried it with Flux so far, not SDXL.
Hello SubtleShader,
Is it normal that I find non-DMD2 workflow a lot more precise and way less repetitive ? You're saying that DMD2 is the best quality workflow but sticking to good old SDXL karras 35 steps works best for me.
With PAG, it's even slower but yields the best quality/adherence
Having said that, big thanks for all the work on those checkpoints, I'm enjoying a lot of XL4 (can't wait to try photo1) To the point I'd test all the biglust 1.6 loras to see if those work with your checkpoint. :)
Thanks. Everyone has a different taste. Let me know about the loras.
Big Love Photo2 Preview
I'm already working on Photo2. It will have these 40 concepts:
360 degree photo, 3d selfie, 69, 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.
Many more new concepts than in Photo1. I'm still tagging and editing. Training should be faster this time.
Thanks for the update. Do you have any plan to finetune Chroma?
@fsdfjn Chroma 1.0 had 600 downloads. Not really popular. Of all the newer models I like Wan best. But also needs too much VRAM and rendering time. So I will stick to SDXL for now.
any plan for pony version sir ? biglove pony 3 was amazing
@adwadcxsamo123 Not really. Looks like Pony2 is still more popular than Pony3. Or XL4 was just more popular. The Big Love Pony versions have cute faces, but are less creative and realistic. XL is more flexible with prompting, Photo1 even more.
@SubtleShader Indeed, not a fan of the plastic look myself. But XL and now PhotoX are amazing, keep it up good sir!
Please, make uvula and throat
@bobbelly404707 Not sure what you mean...
@SubtleShader oh, the uvula and throat concepts are about showing uvula mostly. I've notices Photo1 is not capable of this right now. The concept is mouth open wide and uvula is visible. It is very a common vore fetish. Uvula, tonsils and throat used to work as triggers.
@H4rryM3ss Sorry, not focusing on fetish & BDSM stuff. Best train a lora yourself. Got enough other concepts to still train.
how do i prompt reverse suspended congress. i trying it and cant get it working
Base prompt is anal sex, reverse suspended congress, 1girl, 1boy, penis, holding. Takes some attempts to get these more exotic positions working.
Still getting the best results with Pony3 (probably the most well-behaved Pony/Illu model I've found so far), but Photo1 seems like a big improvement over XL4 imo. Looking forward to Photo2, but really hoping we'll see a Pony4 with the same behavior as Pony3 only with improved photo-realism. Anyway, keep up the good work, appreciate it!
Pony3 does some nice stuff but prompting and realism is way behind Photo1. Pony3 download numbers were also a lot less than XL4. So no really sure if there will be a Pony4.
Love the images & videos that you produce with Photo1. Thanks, guys!
We are a handful of happy naughty guys here... We are happy because of your wonderful BigLove CP, and we are naughty because... we are just guys.
That being said, maybe you should let all the other naughty guys use your photo CP without asking for these 10.000 buzz.
I kinda feel jerking in a desert.
Cheers pal...
@BretChampagne Thanks, nice to hear, pal. Photo1 is free to download now. In case you did not notice. 😁
@SubtleShader
Photo1 is free to download?
This is scandalous ! you owe me 10.000 buzz XD
Is it possible to make a character LORA with this checkpoint?
It is possible with every checkpoint.
If you're going to be using Big Love to generate with the character LoRa you're training, I highly recommend training directly on Photo1. My results doing so have been extremely impressive. The Onetrainer SDXL default settings that Subtle often recommends are a great start if you're unsure of settings to use.
@DrAIvyn where can i find the settings Subtle suggests for training loras?
@Tezozomoctli2 Quoting myself here: "I use the default SDXL lora settings of OneTrainer with Cosine learning, Dora activated and image augmentation deactivated. I always train on the latest Big Love version. I guess the magic happens mostly with the image selection. I'm quite selective, remove all watermarks, correct brightness, contrast & colors if they are off and sometimes crop images. Training images are never smaller than 900 pixels on the larger side. I also avoid low quality and too soft images. Usually 100-250 training images. Only up to 30 images for character loras."
@SubtleShader I made a LORA with 50 images and it went poorly. If you use 30 images, how many repetitions do you make in total? Can I talk to you privately?
@REEZN Just the default settings with repetition left at 1. Sure write me.
I am interested in training a character lora on this as well. I tried with big love xl, using civitai's lora trainer and it looks nothing like my charactar at the end. I will look into onetrainer.
so will the photo1 version work in sdxl workflows interchangeably where xl4 was used? Ive been using xl4 and really like it
@anonameguyman1234252 Hi bro, the same thing happened to me, I'll talk to you.
@anonameguyman1234252 Yep, Photo1 was trained on XL4 and merged with it.
Is it me or it's just not good with fingers, the best looking SDXL model but hands and feet, not so much.
Heard from others that it has the best hands. Also depends on the prompts that you feed it. Can be very subjective. Can vary with different settings, loras, used software etc.
Try with a lower resolution works fine for me
I'm a beginner. I have Forge and I'm not able to create such beautiful images using this model. Can anyone advise?
Please download the DMD2 lora and put it in Forge's lora folder. Drag & drop one of my images onto the txt2img prompt box of Forge and click the ↘️ icon. Then click the Generate button.
Also check the Big Love description for settings and other hints.
Use comfyui it's hard at the begining but best tool when you get it
@perfectgf I know how to use Comfy, but prefer Forge. I only use Comfy for merging.
@SubtleShader Are there things ComfyUI can do with merging models that Forge can't? I've been merging models in Forge for a while and it had given me some pretty nice models, but if Comfy can do stuff like merging LoRA's into models I'll gladly try that instead.
@SubtleShader thank you, I'll try
@jdood Comfy offers more merging options. You just need to find the nodes. And you can merge a dozen models in one workflow and adjust settings without restarting. More convenient.
@SubtleShader great, it works now. I didn't know that lora was so important. thank you.
i tried all prompts even using civitai images but i cant get reverse suspended congress concept to work at all.
Add the Anal Sex lora at https://civitai.com/models/1444070?modelVersionId=1919016
There is a showcase image there.
Maybe its only me but the ultra res workflow with the ultrares lora produces nothing but elongated bodies again. No problems with the other checkpoints. Any updates on workflows or what am i missing? Maybe resolution tips?
Yes, UltraRes produces more distortions with bigger images. Higher lora strength reduced it, but does not eliminate it. It was an experiment. Would need a lot more training, but training takes very long because of the higher resolution.
@SubtleShader Ah i see. Thanks for your work boss. o/
Is there some sort of bias? I try to get a scene with a blonde Swedish male and a black African female -- but as soon as I use blonde or Swedish anywhere, the female will be white with blonde hair and the male always black (except for some very seldom correct pics). Any suggestions? And I tried a lot (loras, break and : separations, even regional prompters etc., as soon as they need to interact, being close, it won't work)
Maybe 1man, 1 african woman or 1man, 1 dark skinned woman. I never tried to prompt male appearance.
Try adding BREAK between the descriptions of both. Should not bleed over then.
@SubtleShader thanksy, but as written - I trued it all. BREAK. 1man: 1boy: 1male: the man: ... Also Regional prompting via that couple froge extension (that works if really forced oc, but won't make them interact right)
@pornthulhu If not even loras work, I have no idea.
@pornthulhu If not even loras work, I have no idea.
@SubtleShader i would need a white male - black female lora. Or something similar. Do you happen to know some loras? I only found "darker skin color" which oc also ends in the male having dark skin. And maybe embeddings for certain characters, but I dont want that :/
Why don t work on fooocus? Don't load
Sorry, no idea. Never used Foocus. Big Love is a standard SDXL model, so it should work.
@SubtleShader when i Copy and paste link in !wget command and click for start ... 0% of download ... And appear immediately that failed authorization ... Please fix It .. i would like to try it
@Rhaenys1 Sorry, I have no influence over Civitai downloads.
any recommendations on how to get rid of the ever present tan lines? I use a dmd2 workflow, do negatives won't work :'(
I will train the concept no tan lines in Photo2.
Try "even skin tone"
@jxucbsxxqehblzal5507 doesnt work for me
Do I need DMD2 for this or can I also gen normally? I see almost all pics on here using DMD2 lora
DMD2 looks much better and is 3-6x faster. XL2.5, XL3, XL4 & Photo1 were optimized for DMD2. You can use without, but not recommend.
@SubtleShader dmd have a disadvantage - all images look more "average". It generates much less unique content even if quality looks better.
@Shio_N Use DMD2 or don't. Your choice.
@SubtleShader Out of curiousity. Could the DMD2 lora be merged into it🤔 or would that not work well?(since you can no longer adjust it)
Since alot of people seem to struggle using it.
@Meower2024 I don't know any people struggling. Anyway, here are the disadvantages of merging the DMD2 lora:
1. Not possible to adjust the DMD2 strength
2. Not possible to generate without it
3. Not possible to use another speed lora
4. Not possible to train a lora on it
5. Not possible to fine-tune it
6. Not possible to merge it normally
7. If DMD2 should be upgraded to DMD3, not possible to upgrade the checkpoint.
So, merging DMD2 into a checkpoint has too many disadvantages. The little bit of conveniance is not worth it. Besides, it is not really any conveniance. I have the DMD2 lora always active as snippet in Forge so it gets automatically added to the prompt.
does anyone find there images give women super crazy long backs? any way to fix this? this is with dmd2 workfflow for xl4, specially if my prompt has someone lying down
Use a small image size than 1024x1496 in txt2img, e.g. 832x1216
ok ty I will try that, is 1024x1496 not supposed to be used in t2i? that is what I was using
@anonameguyman1234252 I use 1024x1496 and like the results. You don't like them apparently. So try smaller.
@anonameguyman1234252 It simple doesnt work. Stay with usual resolutions. I tried some model, which claimed they are perfect with whatever resolution, but no.
@ravonaramis572 I like a lot of the outputs, but I do notice very long bodies often, even in the example images for the model. Not the worst thing ever, but even with prompting it doesnt seem to change anything. It is a little better in the smaller resolution. I am on the xl4 version tho, I need to try the photo version still.
If you mean grotesquely elongated torsos, it's the same with many other checkpoints I've used. Bodies get so long that they have two navels, for example. I've tried a lot of things but I've found no sure way of addressing this problem. I'm just as interested in a solution.
@obakeomo1435 You fix it by using a different image size. Mostly smaller image size, but another aspect ratio can also fix it. SDXL is trained on 1024 pixels. The default size is 832x1216, which always works for txt2img. I'm using 1024x1496 because DMD2 allows that and a majority of images are fine.
@SubtleShader Thanks for the advice.
For example, I had images with torsos so long that there up to 4 navels!
Your fix usually works. A fix I found is to turn off Hires if using Automatic1111 and deal with magnification later.
I'll have to pay more attention to the recommended aspect ratios.
Another weird aspect of AI image generation is that Loras are presented with no description of the behavior or function. You just have to play with them to find out. Even then, the results are often unexpected.
I've been at AI generation for only a few months and have only become aware of some of the subtleties, e.g., aspect ratios. As someone who comes from a software development background, I find it strange that a user should concern himself with details that the system should handle. I guess with time, these issues will be addressed.
@obakeomo1435 is it just a problem with all sdxl stuff it seems? Flux didnt seem to flip out this much in this regard, then again flux can't don nsfw at all.
@obakeomo1435 Sorry, just saw your comment. Never had 4 navels so far. 1024x1496 does not produce that. Loras and checkoints are very versatile and flexible. There is no one way to use them as is the case with software. They are not standard software. They can be used very creatively. So the user needs to find out as he is engaging creatively. I'm giving all recommended settings in the Big Love description, but most don't even care to read & use that.
@obakeomo1435 I've never had 4 navels either. Try the comfy ui workflows on this page here in comfy ui? I've used to use pinokio and wangp with image generation for example and certain models like flux even would give me deformed monstrosities, but using comfy ui workflows for the same thing gave me normal results.
Photo1 - a brilliant checkpoint! 👍
I did not used sdxl for a long time after pony and flux came up. But this checkpoint is really a "game changer" and works perfect with your DMD2 UltraRes Workflow 🙂 I love it! Thank you for your exellent work.
It's really difficult to train a character Lora on this checkpoint, the results are really bad without dmd2, do you have any advice?
The lora works very well with dmd2 though.
I trained against the base sdxl and it works okay and then downloaded a few base sdxl character loras and they seem okay. Curious though what the auther suggests and bonus for any training parameters that can be shared!
Big Love was optimized for DMD2, so why use it without? Train on another checkpoint if you don't want to use DMD2. I posted my lora settings in another comment here.
Thanks for the answer, I only asked because I wanted to know if there was problem on my end.
I'll use the DMD2 lora. Great model btw. @SubtleShader
Hello, can you please tell me why I have very high contrast in all my photos, but I'm using cfg scale-1 and forge.
Only use 8 steps and the LCM sampler. If that is not the problem, prompt low contrast, cloudy or something like that. Alternatively reduce the DMD2 lora strength.
@SubtleShader Thanks for the reply, I'll try.
@Gomane355 Also cfg should be 1 and don't use a lora that increases contrast.
@SubtleShader Yes, I understand that all the settings are the same, but for some reason you have a neutral contrast, unlike mine. Thank you for your answers!
@Gomane355 Post an image with metadata, give me the link and I will check what the problem is.
@SubtleShader thank you for responding and taking your time! I have uploaded the images and attached the link
@Gomane355 Your metadata says "Size: 896x1152, Token merging ratio: 0.7, Discard penultimate sigma: True"
I use 1024x1496 which produces better details. I also don't use these other 2 settings. No idea what they are. Check your Forge settings.
@SubtleShader Okay, thanks, I'll try to figure it out.
amazing checkpoint, can't wait for photo 2, works well with some of my pony loras