Big Love

KreaLens1 (based on KreaEdit1) does image editing &amp; txt2img. It produces sharper images, is more photorealistic &amp; adds more details to images. It was trained on 21 new NSFW/sex concepts and 27 other concepts. It produces cuter faces and can do various body fluids better. You can prompt for skin texture, skin pores or soft skin to modify skin quality.  Improved prompt adherence, e.g. no random tattoos or wetness. See the description for new concepts. Krea Model Lines KreaLens continues the idea of a pro style model that started with KreaEdit1. Future KreaEdit releases will mainly offer enhanced editing capabilities while KreaLife focuses on amateur &amp; candid looks. Variants The bf16 version requires 24 GB VRAM. The 8-bit versions 12 GB VRAM. The 4-bit &amp; GGUF versions only need 8 GB VRAM. The int8 (convrot) version now works with the native Load Diffusion Model node of up-to-date ComfyUI version. To make the GGUF version work in ComfyUI, you need to unzip the third .zip file from Optional Files into the custom_nodes sub folder. Lora Training The training files of KreaLens1 are raw versions. So train it as Krea 2 Raw. Different versions are needed for AI-Toolkit (single raw file) and OneTrainer (diffusers), which can be found under Optional Files.

bigLove_krealens1.gguf

bigLove_krealens1_trainingData_3108685.zip

bigLove_krealens1.zip

bigLove_krealens1_nvfp4.safetensors

bigLove_krealens1_fp8Mixed.safetensors

bigLove_krealens1_trainingData_3110508.zip

bigLove_krealens1_bf16.safetensors

bigLove_krealens1_int8.safetensors

bigLove_krealens1_mxfp8.safetensors

KreaLens1

KreaGlam1

KreaLife2

KreaEdit2

Zora1

Zia2

Gwen3

KreaLife1

KreaEdit1

Klein4

Cat1

Gwen2

Klein3

Gwen1

Eryn1

Ultra5

Photo6

Zuna1

Klein2

Zia1

Hyper1

Klein1

Ultra4

Photo5

Ultra3

Photo4.5

Ultra2

Photo4

Ultra1

Lust1

Photo3

Insta1

Photo2

Photo1

Pony3

XL2.5

Pony2

Pony1

KreaLife2 is much more life-like and photorealistic. The typical krea AI face is gone. It features the new improved editing capabilities of KreaEdit2 and was trained on facial variety, ethnicity, amateur look and 28 new concepts. Prompt adherence was improved.It usually generates average looking faces unless you prompt otherwise (e.g. cute, beautiful, gorgeous). While KreaLife1 produced darker images, KreaLife2 tends to do brighter images. You can prompt against that with e.g. lowkey, dramatic light, dim light etc.The editing results are different from KreaEdit as it applies a different look when creating new scenes via editing. The image style of txt2img and editing is much more similar now. To add more texture to editing results, activate upscaling and use 4-5 steps.New int4 (w4a4 convrot) version that runs fast on most GPUs and only needs 8 GB VRAM. The int8 (convrot) version now works with the native Load Diffusion Model node of up-to-date ComfyUI versions. To make the GGUF version work in ComfyUI, you need to unzip the third .zip file from Optional Files into the custom_nodes sub folder.Improved training files that produce much more precise character loras.

bigLove_krealife2_bf16.safetensors

bigLove_krealife2_mxfp8.safetensors

bigLove_krealife2_int8.safetensors

bigLove_krealife2_int4.safetensors

bigLove_krealife2_trainingData_3247917.zip

bigLove_krealife2_fp8Mixed.safetensors

bigLove_krealife2.gguf

bigLove_krealife2_trainingData_3247553.zip

bigLove_krealife2_nvfp4.safetensors

bigLove_krealife2_trainingData_3248276.zip

Please note: The KreaEdit versions of Big Love are pay-only and special license conditions apply. See description!KreaEdit1 Big Love is 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. It produces only photorealistic images with booru tags (not anime like other Krea 2 models).KreaEdit1 Big Love 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 &amp; fine detail.Move the KreaEdit1 model file into the diffusion_model sub folder of the models folder of ComfyUI, <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Krea-2/resolve/main/text_encoders/qwen3vl_4b_bf16.safetensors?download=true">qwen3vl_4b_bf16.safetensors</a> (12+ GB VRAM) or <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Krea-2/resolve/main/text_encoders/qwen3vl_4b_fp8_scaled.safetensors?download=true">qwen3vl_4b_fp8_scaled.safetensors</a> (6-8 GB VRAM) into the text_encoder sub folder and <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Krea-2/resolve/main/vae/qwen_image_vae.safetensors?download=true">qwen_image_vae.safetensors</a> 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.Editing component by conradlocke called <a rel="ugc" href="https://huggingface.co/conradlocke/krea2-identity-edit">Krea 2 Identity Edit</a>, which allows character restaging, local edits, replacing a person, adding a style and combining scene &amp; character from seperate images. Requires the <a target="_blank" rel="ugc" href="https://github.com/lbouaraba/comfyui-krea2edit/">Krea2Edit nodes</a> that have to be installed manually. For more face consistency try increasing the grounding_px parameter to 1024 or up to 1536.KreaEdit1 needs double as long to render than Klein, Gwen &amp; Cat, but the better results with as good as no anatomy problems are worth it. You can reduce steps to 7, 6 (or even 4) to speed it up, but 8 steps is usually the best quality.Version RecommendationsWith 12+ GB VRAM I recommend using the int8 convrot version (if you have Triton installed in ComfyUI), because it works fastest on all gpus and has the highest 8-bit precision. Otherwise use mxfp8 with 5xxx gpus (if you have cuda 13 installed) or fp8 mixed in all other cases. With 6-8 GB VRAM best grab the nvfp4 version with a 5xxx gpu (if you have cuda 13 installed), otherwise the GGUF version. With 24 GB VRAM you can also use the bf16 version, but needs twice the generation time of int8. Only if you need absolute precision.Lora TrainingThe training files of KreaEdit1 are raw versions. So train it like krea 2 raw. Different versions are needed for AI-Toolkit (single file) and OneTrainer (diffusers). OneTrainer trains double as fast as AI-Toolkit on a 5090.

bigLove_kreaedit1_trainingData.zip

bigLove_kreaedit1_full_nf4.safetensors

bigLove_kreaedit1.gguf

bigLove_kreaedit1_trainingData_3002970.zip

bigLove_kreaedit1_pruned_bf16.safetensors

bigLove_kreaedit1_2994662.safetensors

bigLove_kreaedit1_pruned_fp8.safetensors

bigLove_kreaedit1_2994489.safetensors

bigLove_kreaedit1_trainingData_3002326.zip

bigLove_kreaedit1_trainingData_2995770.zip

KreaEdit2 is more photorealistic, face-consistent, does not resist NSFW editing anymore and offers additional editing capabilities &amp; improvements. It knows 48 new concepts compared to KreaEdit1. KreaEdit2 can do txt2img too. New fast int4 version for 8 GB VRAM.Editing The NSFW editing refusal was decreased from 63% to 9%. Additionally, KreaEdit2 more often only modifies the reference image instead of recreating a similar image im Krea 2 style.With 2 reference images KreaEdit2 can now do person/head/face/eye swap and put clothes on a person. It can do outpainting, inpainting, create and use character reference sheets. Also offers improved object/person removal and local edits.KreaEdit2 offers improved NSFW editing capabilities compared to the identity edit lora. It is recommended to use a grounding_px of 768 for KreaEdit2, because the new editing training was based on it.Variants The bf16 version requires 24 GB VRAM. The 8-bit versions 12 GB VRAM. The 4-bit &amp; GGUF versions only need 8 GB VRAM.New int4 (w4a4 convrot) version that runs fast on most GPUs and only needs 8 GB VRAM. The int8 (convrot) version now works with the native Load Diffusion Model node of a up-to-date ComfyUI version. To make the GGUF version work in ComfyUI, you need to unzip the third .zip file from Optional Files into the custom_nodes sub folder.Lora Training The training files of KreaEdit are raw versions. So train it as Krea 2 Raw. Different versions are needed for AI-Toolkit (single raw file) and OneTrainer (diffusers), which can be found under Optional Files.Other Tools If you want to render 30% faster with minimal quality loss, you can use the <a target="_blank" rel="ugc" href="https://huggingface.co/lvladikov/Krea2-Turbo-Distill-4step-LoRA">4step Turbo lora</a> but with 6 steps.

bigLove_kreaedit2_int4.safetensors

bigLove_kreaedit2_nvfp4.safetensors

bigLove_kreaedit2.gguf

bigLove_kreaedit2_trainingData_3231499.zip

bigLove_kreaedit2_mxfp8.safetensors

bigLove_kreaedit2_int8.safetensors

bigLove_kreaedit2_bf16.safetensors

bigLove_kreaedit2_trainingData_3213176.zip

bigLove_kreaedit2_fp8Mixed.safetensors

bigLove_kreaedit2_trainingData_3230842.zip

bigLove_kreaedit2_trainingData_3212456.zip

bigLove_kreaedit2_trainingData_3212939.zip

KreaGlam1 produces more "glamorous" images, which means rounder faces, more stylish poses, more interesting lighting, more decorative settings etc. It was trained on 34 new concepts (see description!). The editing results are different from KreaEdit &amp; KreaLife as it produces a different look &amp; style when creating new scenes via editing. To add more texture to editing results, activate upscaling and use 4-5 steps. To make the GGUF version work in ComfyUI, you need to unzip the GGUF .zip file from Optional Files into the custom_nodes sub folder. The training files are raw versions. So train it as Krea 2 Raw. Different versions are needed for AI-Toolkit (single raw file) and OneTrainer (diffusers), which can be found under Optional Files.

bigLove_kreaglam1_trainingData_3270364.zip

bigLove_kreaglam1_int8.safetensors

bigLove_kreaglam1_mxfp8.safetensors

bigLove_kreaglam1_trainingData_3270118.zip

bigLove_kreaglam1_int4.safetensors

bigLove_kreaglam1_bf16.safetensors

bigLove_kreaglam1.gguf

bigLove_kreaglam1_nvfp4.safetensors

bigLove_kreaglam1_trainingData_3269942.zip

bigLove_kreaglam1_fp8Mixed.safetensors

Please note: The KreaLife versions of Big Love are pay-only and special license conditions apply. See description!KreaLife1 Big Love is a finetune of Krea 2 that can do txt2img and editing. The focus is more on life-like and less pro-style results. Compared to KreaEdit1 is was trained on 21 new sex concepts &amp; 19 new amateur/candid concepts. See the description for concept names!It achieves higher face consistency (provided you update the <a target="_blank" rel="ugc" href="https://github.com/lbouaraba/comfyui-krea2edit">Krea2Edit nodes</a> to the latest version). Use a ref_boost value of 3 or 4.Version RecommendationsWith 12+ GB VRAM I recommend using the int8 convrot version (if you have Triton installed in ComfyUI), because it works fastest on all gpus and has the highest 8-bit precision. Otherwise use mxfp8 with 5xxx gpus (if you have cuda 13 installed) or fp8 mixed in all other cases. With 6-8 GB VRAM best grab the nvfp4 version with a 5xxx gpu (if you have cuda 13 installed), otherwise the GGUF version. 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. With 24 GB VRAM you can also use the bf16 version, but needs twice the generation time of int8. Only if you need absolute precision.Lora TrainingThe training files of KreaLife1 are raw versions. So train it like krea 2 raw. Different versions are needed for AI-Toolkit (single file) and OneTrainer (diffusers). OneTrainer trains double as fast as AI-Toolkit on a 5090.

Files with SHA256: 881c4dbde2e83ef4b2b1aeefcc4773eb72733c73b5cfe99ec7ec3dae3f2cf9cb