vSFW unless prompted otherwise ( !!! WARNING ⚠️ NO SAFETY FILTER !!! ) bypass any LoRAs you don't want to use
Runtime profile
Source description
vSFW unless prompted otherwise ( !!! WARNING ⚠️ NO SAFETY FILTER !!! ) bypass any LoRAs you don't want to use
models:
diffusion_models: https://huggingface.co/Comfy-Org/Krea-2/blob/main/diffusion_models/krea2_turbo_int8_convrot.safetensors
text_encoders: https://huggingface.co/Winnougan/Krea-2-Base-Turbo-NVFP4-FP8-INT8/blob/main/Text%20Encoder/qwen3vl_4b_int8.safetensors
AI-generated commentary
AI-generated explanation based on source and configuration details. Suggestions are clearly labeled.
Krea 2+ is a Krea 2 image-generation workflow: enter a text prompt and use its model files to generate an image for preview or saving.
The workflow encodes the prompt, samples a latent image, decodes it with a VAE, then previews or saves it.
The listed model files are krea2_turbo_int8_convrot.safetensors and qwen3vl_4b_int8.safetensors, plus either wan_2.1_vae.safetensors or qwen_image_vae.safetensors.
Images can be previewed and saved.
This is version v3 of Krea 2+, published by L4y3r.
No license is listed.
Other named model files include galaxyace_krea2.safetensors, KNP_V2_copy_copy.safetensors, Krea-2-Turbo-Projector-Scale-LoRA-Diffusers.safetensors, krea2_Enhancer.safetensors, Krea2_Nikki_Style_Anime_000006000.safetensors, krea2_turbo_fp8.safetensors, and krea2-mikkoph.safetऱs.
Also named are Krea2-realism-V1.safetensors, krea2filterbypass3.safetensors, qwen_image_vae.safetensors, qwen3vl_4b_bf16.safetensors, redcraftKREA2RedMix_krea2Edition.safetensors, SummerVibesHM_krea2_epoch8.safetensors, and syn4pse_krea2.safetensors.
To leave out a LoRA, right-click its LoRA Loader node and choose Bypass.
Suggestion · not verified
The notes say you can swap in other LoRAs and adjust their strengths.
Suggestion · not verified
Krea 2+ is described as vSFW unless prompted otherwise, and the notes say it has no safety filter.
Does this need to be edited?
Sign in to send an edit request.
Sources
1 sourceSource excerpts
3 excerptsSource context: 1366 downloads · Type Workflows · Base model Krea 2
OR
loras:
lenovo_krea2: https://civitai.com/models/1662740/lenovo-ultrareal
nicegirls_krea2: https://civitai.com/models/1862761/nicegirls-ultrareal
more loras: [optional]
if you don't want to download all of these you don't have to, just bypass the errored nodes (red outline when you hit run and it says missing models) — right click on LoRA Loader Node then select Bypass for each node you don't use
even more loras:
these aren't in the workflow but you can feel free to un-bypass and swap out any LoRAs you wish and adjust the strength of them as you see fit. The organization of the workflow groups is quite bad because I am profoundly retarded, but it helps keep everything on the screen with a large font at 111% zoom in 1440p 😎
SummerVibesHM_krea2_epoch8: https://civitai.com/models/2727284/realism-enhancer-krea2
Krea2_Nikki_Style_Anime_000006000: https://civitai.com/models/2727496/nikke-goddess-of-victory-style-krea2?modelVersionId=3065893
Screenshots:
v3:
v2:
v1:
nsfw:
make sure you bypass the node and also the switch next to it, switch boolean set to true will be moody, false is the default Krea 2 Turbo INT8 — i bypass these nodes cause most people won't use moodyMixv3 and don't want their ComfyUI to show an error for another useless missing LoRA!
(depreciated v5):
Must (un)Bypass BOTH the Model [UNET Loader] AND the switch to enable it, then you can toggle between SFW and NSFW using the switch — true = NSFW / false = SFW
[replaces krea2_turbo_fp8 using unet switch — right click & Bypass to enable]
Abliterated clip isn't necessary unless you really just aren't getting the actions you're describing regardless of how you prompt for it.
[replaces qwen3vl_4b_bf16 — abliterated produces more graphic content]
⚠️
Estimated VRAM requirement
Estimate unavailable
1.80 GB across 4 of 17 model files. Model file total + 25% loading overhead + 2 GB execution buffer, rounded up.
Requirements
27 requirementsCLIPLoader
PossibleText encoder · Unknown
Checkpoint · 109 MB · Hugging Face · saik0s/comfy_backup
KNP_V2_copy_copy.safetensors
Not resolvedLoRA · Unknown
Checkpoint · 268 B · Hugging Face · Beinsezii/Krea-2-Turbo-Projector-Scale-LoRA-Diffusers
Checkpoint · Hugging Face · vrgamedevgirl84/Krea2_Enhancer · Krea2 Enhancer
Krea2_Nikki_Style_Anime_000006000.safetensors
Not resolvedCheckpoint · Unknown
Checkpoint · Civitai · 958009
Checkpoint · Civitai · 2415137
Checkpoint · 1.46 GB · Civitai · 2728365
krea2filterbypass3.safetensors
File unverifiedCheckpoint · Hugging Face · uzumix/krea2filterbypass3.safetensors · Krea2filterbypass3.safetensors
VAE · 242 MB · Hugging Face · Comfy-Org/Krea-2
Text encoder · Hugging Face · ahmed22xa/Huihui-Qwen3-VL-4B-Instruct-abliterated-comfy · Huihui Qwen3 VL 4B Instruct Abliterated Comfy
redcraftKREA2RedMix_krea2Edition.safetensors
Not resolvedCheckpoint · Unknown
SummerVibesHM_krea2_epoch8.safetensors
Not resolvedCheckpoint · Civitai · 2727284
Checkpoint · Civitai · 2734426
VAEDecode
PossibleVAE · Unknown
VAELoader
PossibleVAE · Unknown
CLIPTextEncode
PossibleNode pack · Unknown
ComfySwitchNode
Not resolvedNode pack · Unknown
ConditioningZeroOut
PossibleNode pack · Unknown
EmptyLatentImage
Source context: 1246 downloads · Type Workflows · Base model Krea 2
Source context: 1137 downloads · Type Workflows · Base model Krea 2
Node pack · Unknown
KSampler
PossibleNode pack · Unknown
LoraLoader
PossibleNode pack · Unknown
LoraLoaderModelOnly
PossibleNode pack · Unknown
PreviewImage
PossibleNode pack · Unknown
SaveImage
PossibleNode pack · Unknown
UNETLoader
PossibleNode pack · Unknown