--- tags: - text-to-image - lora - diffusers - template:diffusion-lora widget: - output: url: images/0.png text: 'face headshot, a Middle Eastern male with thick black hair styled in a fade and a well-groomed beard. He’s wearing a fitted white turtleneck sweater. His lips are rose-tinted, and his almond-shaped brown eyes are calm but confident. The backdrop is a deep wine red, adding richness to the composition.' - output: url: images/1.png text: 'face headshot, a woman with long brown hair, wearing a gray turtleneck, is stunning against a vibrant red wall. Her left hand is resting on her chin, adding a touch of warmth to her face. The womans gaze is directed to the right, her lips are pursed, revealing a soft pink lip. Her eyebrows are a darker shade of brown, and her hair cascades down to her shoulders. The wall behind her is a soft shade of red, and the womans hair is pulled back in a ponytail.' - output: url: images/11.png text: 'face headshot, High-resolution photograph, woman, UHD, photorealistic, shot on a Sony A7III --chaos 20 --ar 1:2 --style raw --stylize 250' - output: url: images/2.png text: 'face headshot, an African American woman with short curly black hair, wearing a yellow off-shoulder top. Her lips are a glossy shade of red, and her eyes are warm brown, reflecting a sense of calm. She has gold hoop earrings, and her skin glows under soft studio lighting. The backdrop is a muted teal, enhancing the radiance of her complexion.' - output: url: images/3.png text: 'face headshot, a Caucasian woman with shoulder-length blonde hair styled in loose waves, wearing a navy blue blazer over a white blouse. Her lips are nude with a soft matte finish, and her hazel eyes are accentuated with subtle eyeliner. She is smiling gently, revealing her warmth. The background is a gradient of soft gray, giving the portrait a professional tone.' - output: url: images/4.png text: 'face headshot, an East Asian man with medium-length black hair styled messily, wearing a black leather jacket over a plain white t-shirt. His lips are pale pink, and his eyes are a sharp gray-brown, looking away from the camera. The background is urban-inspired, a textured concrete gray wall, giving the portrait a cinematic edge.' - output: url: images/5.png text: 'face headshot, a Latina woman with long straight black hair, parted in the middle and tucked behind her ears. She is wearing a red satin blouse with subtle shimmer. Her lips are painted a bold crimson, and her almond-shaped eyes are dark brown, framed with long lashes. The backdrop is soft golden beige, illuminating her features.' - output: url: images/6.png text: 'face headshot, a Caucasian male with wavy blonde hair swept back, wearing a white linen shirt slightly unbuttoned at the collar. His lips are pale pink, and his piercing blue eyes are directed slightly upward, as if lost in thought. The background is a pale sandy beige, evoking a beach-inspired vibe.' - output: url: images/7.png text: 'face headshot, a Hispanic male with close-cropped hair and a full beard, wearing a black crew-neck t-shirt. His dark brown eyes are intense, and his lips are neutral with a serious expression. The background is a dramatic dark gray, adding depth to the mood.' base_model: Qwen/Qwen-Image instance_prompt: face headshot license: apache-2.0 --- ![1.png](https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/u9flSy6mYiIw8nGckwWVg.png) # Qwen-Image-HeadshotX > [!note] **Qwen-Image-HeadshotX** is a super-realistic headshot adapter for [Qwen-Image](https://huggingface.co/Qwen/Qwen-Image), an image generation model by Qwen. It is an advanced LoRA adaptation of the Qwen-Image model and an upgraded version of [Qwen-Image-Studio-Realism](https://huggingface.co/prithivMLmods/Qwen-Image-Studio-Realism), offering more precise portrait rendering with a strong focus on realism. The model was trained on diverse face types from across the world, labeled with **florence2-en** and caption-optimized using [DeepCaption-VLA-7B](https://huggingface.co/prithivMLmods/DeepCaption-VLA-7B). **Total Images Used for Training:** 55 RAW images \[11(types) × 5 different face types: Asian, Hispanic, Caucasian, Latina, Middle Eastern, etc.]. --- # Model description for Qwen-Image-HeadshotX Image Processing Parameters | Parameter | Value | Parameter | Value | |---------------------------|--------|---------------------------|--------| | LR Scheduler | constant | Noise Offset | 0.03 | | Optimizer | AdamW | Multires Noise Discount | 0.1 | | Network Dim | 64 | Multires Noise Iterations | 10 | | Network Alpha | 32 | Repeat & Steps | 25 & 4000 | | Epoch | 30 | Save Every N Epochs | 1 | Labeling: florence2-en(natural language & English) + 🔥 Optimized with Long-Caption VLA Multimodal : https://huggingface.co/prithivMLmods/DeepCaption-VLA-7B Total Images Used for Training: 55 RAW [11(types)×5 different face types (Asian, Hispanic, Caucasian, Latina, Middle Eastern, etc.)] ## Data Sources | Source | Link | |--------------|-------------------------------------| | Playground | [playground.com](https://playground.com/) | | ArtStation | [artstation.com](https://www.artstation.com/) | | 4K Wallpapers| [4kwallpapers.com](https://4kwallpapers.com/) | ## Best Dimensions & Inference | **Dimensions** | **Aspect Ratio** | **Recommendation** | |-----------------|------------------|---------------------------| | 1472 x 1140 | 4:3 (approx.) | Best | | 1024 x 1024 | 1:1 | Default | ### Inference Range - **Recommended Inference Steps:** 45-50 (approx. ~ `100 Seconds Inference`) ## Setting Up ```python import torch from diffusers import DiffusionPipeline base_model = "Qwen/Qwen-Image" pipe = DiffusionPipeline.from_pretrained(base_model, torch_dtype=torch.bfloat16) lora_repo = "prithivMLmods/Qwen-Image-HeadshotX" trigger_word = "face headshot" pipe.load_lora_weights(lora_repo) device = torch.device("cuda") pipe.to(device) ``` ## Trigger words You should use `face headshot` to trigger the image generation. ## Download model [Download](/prithivMLmods/Qwen-Image-HeadshotX/tree/main) them in the Files & versions tab.