Big Love

Big Love Qiara1 is a finetune of Qwen Image 2.1 for image editing and txt2img. It is reproduces faces very reliably and anatomy is very good. Qiara1 is just as tuck sharp as the base model. It was trained on various NSFW concepts giving it a Western look. For best quality render at least 2 megapixels (1184x1776) or even 4 megapixels (1664x2512). The workflow includes the custom sampler with official sigmas (extended for 6 &amp; 10 steps) and a KSampler, which can be used as an alternative. There is also a deactivated option to adjust the softness/sharpness by modifying a turbo lora ratio (between -0.3 to 0.5). To use it you need to download the turbo lora file under Optional Files and place it in ComfyUI's lora folder. For more sharpness you can also use the er_sde sampler instead of Euler. You can use my Detailer workflow to enhance details (see description). To make the GGUF version work in ComfyUI, you need to install and add the <a target="_blank" rel="ugc" href="https://github.com/city96/ComfyUI-GGUF">ComfyUI-GGUF</a> node. Make sure you update ComfyUI for Qwen Image 2.1 support. To run Qiara1 you need to copy  <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Qwen-Image-2.1/resolve/main/text_encoders/qwen3vl_8b_w4a8.safetensors?download=true">qwen3vl_8b_w4a8.safetensors</a> (8 GB VRAM) or <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Qwen-Image-2.1/resolve/main/text_encoders/qwen3vl_8b_int8_convrot.safetensors?download=true">qwen3vl_8b_int8_convrot.safetensors</a> (12+ GB VRAM) into the text_encoder sub folder and <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Qwen-Image-2.1/resolve/main/vae/qwen_image_2.1_vae_bf16.safetensors?download=true">qwen_image_2.1_vae_bf16.safetensors</a> into the VAE sub folder of the models folder of ComfyUI. Big Love Qiara1 is released under the <a target="_blank" rel="ugc" href="https://github.com/QwenLM/Qwen-Image-2.1/blob/main/LICENSE">Qwen Research License</a>, but Qwen grants a (Free Tier) Commercial License for free, if you generate less than 300k images per month. Just email <a target="_blank" rel="ugc" href="mailto:model-business@notice.qwencloud.com">model-business@notice.qwencloud.com</a> (as mentioned in the license), describe your use case, state your amount of images per month and your country. No training files are provided, because it is recommended to train with the turbo option of AI-Toolkit. You can train directly on the int8 or bf16 version of Qiara1.

BigLoveQiara1-Q5_K_M.gguf

BigLoveQiara1_fp8mixed.safetensors

BigLoveQiara1_bf16.safetensors

BigLoveQiara1_nvfp4.safetensors

BigLoveQiara1_int4.safetensors

BigLoveQiara1_int8.safetensors

BigLoveQiara1_mxfp8.safetensors

qwen21_turbo_rank256.safetensors

Qiara1

KreaEdit3

KreaGlam1

KreaLife2

KreaEdit2

Zora1

Zia2

KreaLens1

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

Files with SHA256: 2f49a17e30819c045391beb66a728d2b1ebcbb28f5a370447d94f6db23b2b2ef