You can get KreaLife1, KreaEdit1, Klein4, Cat1, Gwen2 or Hyper1 for $10 and Ultra5, Zuna1 or Zia1 for $5.50 on Ko-Fi. 30-day membership subscriptions available there too. 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.
KreaEdit/KreaLife Big Love are a finetune of Krea 2 that can do txt2img and NSFW image editing. It is more photorealistic, a lot more NSFW-capable and has better skin texture than the original Krea 2. They produce only photorealistic images with booru tags (not anime like other Krea 2 models). They offers the best anatomy (with the least problems) of all Big Love checkpoints. It was optimized for 8 steps with er_sde sampler and beta scheduler for increased sharpness & fine detail.
Move the model file into the diffusion_model sub folder of the models folder of ComfyUI, qwen3vl_4b_bf16.safetensors (12+ GB VRAM) or qwen3vl_4b_fp8_scaled.safetensors (6-8 GB VRAM) into the text_encoder sub folder and qwen_image_vae.safetensors into the vae sub folder. To make the GGUF version work in ComfyUI, you need to unzip the second .zip file from Optional Files into the custom_nodes sub folder.
For editing they require the Krea2Edit nodes that can be installed via Comfy Manager by searching for "Krea 2 Identity Edit". For more face consistency try increasing the grounding_px parameter to 1024 and ref_boost to 3 or 4.
Settings are 8 steps (3-4 steps for upscaling), cfg1, er_sde sampler, Beta scheduler, 1024x1536 pixels. Drag'n'drop showcase images into ComfyUI for a workflow.
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. Check out these anatomy tips.
Settings are 4 steps (2 steps for upscaling), cfg1, Euler sampler, Beta or Simple scheduler, 832x1216, 1024x1536, 1280x1920 or 1536x2240 pixels. Drag'n'drop showcase images into ComfyUI for a workflow.
Big Love Cat is a finetune of LongCat and allows NSFW image editing. It reproduces faces extremely well, with almost every image, so a lot more reliable than Klein & Gwen. It is only 6B and needs 8 steps cfg 1. Same Apache 2.0 license and text encoder as Qwen Image but 8 GB VRAM are enough. Txt2img and sex scenes are still experimental. The base model was only trained on one input image & only picks up random details from a 2nd image. The bf16 version of Cat is faster than Klein & Gwen. The mxfp8 version is 1.6x faster than bf16 and just as fast as Gwen.
There is a new detailer workflow below that I developed for Cat but can be used with any model's images. It uses Photo, Ultra, Hyper or Klein for detailing nipples and pussies.
It is recommended to use ComfyUI. Move the 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 ae.safetensors into the VAE sub folder.
Training speed is more than 2x faster than any transformer model that I know. For training a lora you need SimpleTuner, which requires WSL under Windows and has a WebUI similar to AI-Toolkit. Detailed instructions will become available soon...
Big Love Cat produces the sharpest images with er_sde/Normal. In a few cases this adds subtle color grain which can be avoided by increasing steps to 12. er_sde/beta and Euler produce softer images without grain.
Settings are 8-12 steps (4 steps for upscaling), cfg 1, er_sde/Normal sampler/scheduler (Softer: Euler/Beta or er_sde/Beta), 832x1216, 1024x1536 or 1232x1824 pixels. Drag'n'drop showcase images into ComfyUI for a workflow.
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 Gwen allows commercial usage without additional fees (if no server generation is 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 Gwen workflows offer a skin enhance feature that is not available in Forge Neo. Move the Gwen 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.
Settings are 6-8 steps (3 steps for upscaling), cfg 1-1.5, Euler sampler, Beta scheduler, 832x1216 or 1024x1536 pixels. Drag'n'drop showcase images into ComfyUI for a workflow.
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. Drag'n'drop showcase images into ComfyUI for a workflow.
Big Love Hyper supports generating at image sizes of 1664x2432 (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. Look below for a ComfyUI workflow.
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. Here is a ComfyUI workflow embedded in a image.
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. Look below for a ComfyUI workflow.
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. Look below for a ComfyUI workflow.
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. Look below for a ComfyUI workflow.
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.
Generation Workflows
Download one of the following images and drag'n'drop it onto ComfyUI:
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
Workflows for all other models are embedded in the showcase images above.
Detailer Workflows
Here are the detailer workflows that I use myself. Drag'n'drop the images linked below into Comfy:
Photo/Ultra/Hyper Detailer Workflow
Klein Detailer Workflow
Check the note at the top of the workflow for downloading the necessary detection models and where to place them.
Krea 2 Trained Concepts
KreaLife1: amateur photo, outdoor sex, party, real life, candid, candid shot, outdoor flashing, public indecency, public nudity, beach life, 3d selfie
film grain, low image quality, low contrast, flash light, blurry, soft skin, overexposure
69 position, anal gape, cumshot, dildo insertion, double oral, handjob, innie pussy, object insertion, anal fingering, cuddling, cum swapping, female masturbation, finger licking, footjob, french kiss, funny tongue out, orgasm face, penis licking, pussy licking, vaginal fingering, partner fingering
SDLX 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), Big Love Z-Image, Ernie, Qwen & Longcat versions are licensed under Apache 2.0, Big Love Klein versions are licensed under the FLUX Non-Commercial License v2, and Krea 2 versions of Big Love are licensed under the KREA 2 Community License with the following additional terms:
For clarity, the following version tiers are defined:
Restricted versions: Ultra, Hyper, Klein, Zia, Zuna, Eryn, Gwen, Cat, KreaEdit, KreaLife
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, Gwen, Cat, KreaEdit and KreaLife 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
Please note: The Gwen versions of Big Love are pay-only and special license conditions apply. See description!
Big Love Gwen2 produces better skin texture, improved genitals and higher image quality without artifacts. Sex scenes work more reliable now, especially if you provide details. tx2img and upscaling are also a lot better.
For editing you can now use booru tags (just avoid identity tags like black wavy hair, blue eyes, high cheekbones, slim, curvy, large breasts) or a mix of natural language and booru. txt2img prompts can be used for editing after removing identity tags.
New mxfp8 version that is 1.66x faster than fp8 mixed and much more precise (so less anatomy problems). Works with 16 GB VRAM (better 24+ GB VRAM) with 5060ti and higher.
Smaller nvfp4 version with much higher quality that fits into 16 GB VRAM on 5060ti and higher.
The new nf4 version (for 12+ GB VRAM) works reliably now and produces higher quality despite smaller size. Better than the GGUF version, which is bigger and has lower quality.
Big Love Gwen2 mxfp8 only takes 5 seconds for editing an 1024x1536 image on a RTX 5090. It took 9-10 seconds) with Gwen1 fp8 mixed.
Available Versions:
1. fp8 (19.6 GB): BigLoveGwen2_mxfp8.safetensors (for RTX 5060ti and higher or AMD CDNA3/RDNA4 gpus, 16+ GB VRAM, 1.66x faster than fp8 mixed, almost bf16 precision)
2. fp8 pruned (19.1 GB): BigLoveGwen2_fp8mixed.safetensors (less precise than mxfp8, more precise than nvfp4, 16+ GB VRAM)
3. nf4 (13.5 GB): BigLoveGwen2_nvfp4.safetensors (RTX 5060ti and higher, 1.66x faster than fp8 mixed, slow on older gpus):
4. nf4 Pruned (10.3 GB): BigLoveGwen2_nf4.safetensors (12+ GB VRAM, faster & better quality than GGUF, fast on older gpus):
5. GGUF (13.1 GB): BigLoveGwen2-Q5_K_M.gguf (for 12+ GB VRAM, only if above versions are too slow or don't work)
Optional Files:
1. BigLoveGwen2_diffusers.zip (for training)
2. ai-toolkit_mod.zip (AI-Toolkit patch for training with 32 GB VRAM)
Checkout my Gwen Training comment about training a lora with AI-Toolkit.
FAQ
Comments (39)
Gwen1 model, so far only resulted in black images. I don't know what I am doing wrong. I used the embeded workflow with the exemples images. Updated ComfyUI and all its nodes... still not able to generate something else than a black image. Process seems to run error-free (no message for a problem at a given node). Anyone faced the same issue and would be willing to share knowledge to solve my problem ? Thanks in advance... CDLV
Try downloading the model again. Might have been corrupted during download. Download a version that fits in your VRAM. Try other Gwen workflows.
@SubtleShader I'll test that ! I tried the three possible branching for the models in the workflow from your exemples. So far, none resulted in anything else than a black picture. I'll DL the whole set again and will test it
@charlesdelavigere743 Good luck. I know how annoying this can be. Never heard about such a problem.
Are you using sageattention?
@coyotte He uses my workflow. No sage attention in it. Or is that a Comfy setting?
@SubtleShader --use_sage-attention flag in the .bat of launcher, if you have it installed.
@coyotte I used the provided workflow of SubtleShader. Redownloaded the whole set.. still stuck with black images only... I'll try the sageattention but don't have time to explore this now...
@charlesdelavigere743 Maybe do a second ComfyUI installation. I use three installations of Comfy myself.
@charlesdelavigere743 no, what I meant is that sageattention causes that on some models.
@coyotte So he should deactivate the sage attention Comfy startup arg.
@SubtleShader The week end is coming and will give me some time to test this sageattention approach. I'll keep you posted. Took several attempts to load the models btw, it kept stopping in the middle of every DL ! (probably an issue on my side and my Internet connection...). I only tested those where no issue happened during the whole DL process.
@charlesdelavigere743 So my initial suspicion that your download might have been corrupted might be right. Other than that I recommend a fresh Comfy installation.
@SubtleShader I don't know, as I DL everything once again and it did not solve the issue. I am aiming at the sageattention or find a different way to start comfyui (currently I start it using UmeAiRT-Start-ComfyUI)
Turning off the sageattention in the .bat fixed the problem ! Thank you Coyotte !!!
@charlesdelavigere743 Great to hear!
And great to play around with this amazing model... it was worth digging to the solution !
any idea where I could set the resulting image size in your embedded workflow ? I am sure it's right in front of my eyes...but... well... how to say that without admitting I didn't find it :-) ?
@charlesdelavigere743 There is a scale node that has a megapixel parameter. It automatically keeps the aspect ratio of the input image.
@SubtleShader Do you plan on making a Anima-based model of Big Love?
It’s advertised as an anime model, but there are already some realistic finetunes . The prompt adherence is insane
The license is not great. You basically give the rights to your work to the guys behind Anima and everyone who pays them. It is not very photorealistic. I saw realistic finetunes and they were worse than SD1.5. Prompt adherence might be great for anime stuff, likely poor for realistic. Sorry, I won't train models that I dislike.
I didn’t know the licensing was that bad. Thanks for the response.
Great work!, can you please make an Ideogram version with the new model? I'd love to support it anyway I can.
Please don't ask me that! It is not very photorealistic, it is censored, has a non-commercial license. Ideogram will likely not allow commercial usage of NSFW images either. All red flags. One of these models that are DOA.
@SubtleShader Actually, the censorship there is not worse than in klein or z-image, but the first one to release a decent NSFW model will attract a good audience ;-)
Update: When going through the json conversion process no prompts I tried triggered the censorship filter and it works surprisingly well when you adjust the bboxes to your exact composition needs.
So I'll gladly eat my words here. The statement that regular non-json prompts allegedly work is the lie that tripped the filter for every prompt I tried.
@LbhMan The censorship is on another level in Ideogram and the tedium of having to go through a json conversion process first is just extra inconvenience on top.
While using regular prompts on it triggers the censorship on pretty much anything, like literally "a cat"...
That's as DOA as you can get 😂
@RedPinkRetro if you simply use a node in Comfy or an HTML tool to build your JSON you'll never get a censorship warning for anything again. It just ... doesn't know anything sexual. The "censorship" warning should actually say "Prompt not in JSON format" because if you simply add two or three elements to the JSON object it will never fail.
@RedPinkRetro I saw that the ideogram censorship was bypassed within a couple of days, so I don't think it's serious.
You need to pay for a commercial license to get the bf16 weights. No one else ever did this before. Another problem is that it needs two models (normal and unconditioned) for best quality.
@SubtleShader Thank you for your answers, I will wait, I hope the situation will change in the future.
Does mxfp8 work on 4090rtx?
Sure, it works. No idea how fast. It is very fast on 5xxxx.
Firstly, i've paid a fair bit for these. Without regrets. So obviously I like where this is going and the obvious skill behind its creation. I've been using the latest klein 9b model, and while I love how you can create amazing edits , there's this "ringlight/cinematic/bright/sharp" thing going on that does not quite land in realistic territory. To those of you with this problem, i suggest prompting with "add slight blur to image" and use Claude to remove all forms of cinematic lighting possibilities. The upside of this model is that it can create very specific and detailed outputs. It feels refined and precise. The downside, that cinematic bright light. It's better than waxy/ai look, but it's still hard to have anything look photo real, even with good prompting skills. In the future, i'd love to see versions of this model that focus on a softer more realistic photo look. But with all of the other aspects intact. I think that would be a real hit for those of us that like realism. (I find that real dream is a good example of this)
I already did some training to reduce the bright look but seems it came back a bit with Klein3.Will reduce it again with Klein4. But you can also prompt other types of lighting (e g. soft light, soft shadow) to avoid it.
You can use less steps to make images look softer, by the way. The more steps you use, the more texture it adds.Other than that you can always train your own lora to produce your desired style.
"add slight blur to image" is a really good prompt trick for klein, there are several iterations of it that will improve the results (like focusing on the subject or parts of the background, and describing the blur). I find F2K is the best model we have right now, but so hard to prompt sometimes. Feels like those ASP dark days of SDXL.
@makiaeveli definitely advantages to these models vs what I've ever seen in SDXL though in terms of potentiality of situations and editing in on another level. Thanks for the "type" of blur suggestion, I haven't gone there yet.
@SubtleShader I'll try those suggestions. But I absolutely love something to a bit more of a organic baseline, i don't know if you've ever used chroma but I was a huge fan of how it made things have this hyper organic look when it did not break. Totally different technology and datasets, i know. But natural (not imposed) softness can do wonders for reproducing accurate detail, because what's in biglove in terms of what it's able to accomplish outside of that, would be perfect with neutral details.
@whitespider9999 Never was a fan of Chroma and decided against doing a Big Love version of it. But I know what you mean. My SDXL models still add more fine realistic details than all the newer models, so always good to use them for img2img or detailing. Will still take some time to get Klein, Qwen and Z to that level.
@SubtleShader i found chroma painfully broken in many areas, but it could do emotions and details of human expression a bit more like how you actually see them when you got it all lined up. It had little glimmers of non-souless eyes.