<h3 id="instara-presents:-1girl-qwen-image-v3">Instara Presents: 1GIRL QWEN-IMAGE V3</h3>Join the <a target="_blank" rel="ugc" href="https://discord.gg/instara">Instara Discord</a> for direct support, advanced workflows, and exclusive access to the automatic prompt generator used to generate the prompts for the showcase images. + bonus Gdrive link to huge pack of images full of metadata (workflows+prompts incl.). Check the #announcements channel for these resources. Best workflows for this LoRa: <a target="_blank" rel="ugc" href="https://pastebin.com/rggzbZej">1GIRL (STAND-ALONE) V3</a> <a target="_blank" rel="ugc" href="https://pastebin.com/ymJX5sKt">1GIRL (+SAMSUNG by Danrisi) V3</a> Download the dependancies needed to run Qwen-Image (target paths shown in link text): <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_bf16.safetensors">ComfyUI/split_files/diffusion_models/qwen_image_bf16.safetensors</a> <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b.safetensors">ComfyUI/models/text_encoders/qwen_2.5_vl_7b.safetensors</a> <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors">ComfyUI/models/vae/qwen_image_vae.safetensors</a> <a target="_blank" rel="ugc" href="https://civarchive.com/api/download/models/2270374?type=Model&amp;format=SafeTensor&amp;token=00d790b1d7a9934acb89ef729d04c75a">ComfyUI/models/loras/samsungcam.safetensors</a> BIG SHOUTOUT TO THE 𓃵 GOAT @Danrisi for his amazing support &amp; LoRas. Go check them out now - that is if you know what's good for ya... (•̀ᴗ•́ )و<h3 id="model-information:1girl-qwen-is-a-specialized-lora-fine-tuned-for-the-qwen-image-model.-it-is-engineered-to-replicate-the-authentic-often-flawed-and-candid-aesthetic-of-amateur-smartphone-photography.">Model Information: 1GIRL QWEN is a specialized LoRA fine-tuned for the Qwen-Image model. It is engineered to replicate the authentic, often flawed, and candid aesthetic of amateur smartphone photography.</h3>This model is the result of a meticulous training process on a curated dataset of 1111 images, each with a state-of-the-art descriptive caption. It excels at creating "in-the-moment" scenes that feel captured for social media - raw, unpolished, and hyper-realistic in their imperfections.<hr /><h3 id="what's-included-and-technical-notes">What's Included &amp; Technical Notes</h3>This is a resource-intensive model designed for high-end hardware. For optimal performance, we recommend using a cloud instance (e.g., H100), where generation time averages ~120-150 seconds (1440x1920 reso).Download the LoRa file from Civitai:<ul><li>The LoRA File: <code>1GIRL_QWEN_V1.safetensors</code>. (Click download button)</li></ul>Download Workflows from Pastebin:<ul><li><a target="_blank" rel="ugc" href="https://pastebin.com/rggzbZej">1GIRL (STAND-ALONE) V3</a></li><li><a target="_blank" rel="ugc" href="https://pastebin.com/ymJX5sKt">1GIRL (+SAMSUNG by Danrisi) V3</a></li></ul>Download the dependancies needed to run Qwen-Image (target paths shown in link text): <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/diffusion_models/qwen_image_bf16.safetensors">ComfyUI/split_files/diffusion_models/qwen_image_bf16.safetensors</a> <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/text_encoders/qwen_2.5_vl_7b.safetensors">ComfyUI/models/text_encoders/qwen_2.5_vl_7b.safetensors</a> <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/resolve/main/split_files/vae/qwen_image_vae.safetensors">ComfyUI/models/vae/qwen_image_vae.safetensors</a> <a target="_blank" rel="ugc" href="https://civarchive.com/api/download/models/2270374?type=Model&amp;format=SafeTensor&amp;token=00d790b1d7a9934acb89ef729d04c75a">ComfyUI/models/loras/samsungcam.safetensors</a><hr /><h3 id="the-definitive-guide-to-the-1girl-qwen-v3.0">The Definitive Guide to the 1GIRL QWEN V3.0</h3>Mastering the Art of Flawed Authenticity for Qwen ImageWelcome to the <code>1GIRL QWEN</code> LoRa. This model is not designed for sterile, photorealistic perfection. It is a finely-tuned instrument for capturing the "amateur gaze" - the unique, flawed, and authentic aesthetic of photos taken on smartphones. To get world-class results, you must learn to think like a photographer's friend, not a studio professional.The trigger word is <code>1girl</code>.<hr /><h3 id="the-two-core-styles:-your-creative-choice">The Two Core Styles: Your Creative Choice</h3>The LoRa excels at two distinct types of amateur captures. Choosing your style is the most important decision you'll make.STYLE 1: The IG Post LookThis style mimics a clearer, more intentional photo taken with a modern smartphone for an Instagram post. The moment feels captured on purpose. The imperfections are related to lighting and the limitations of a phone sensor, not video compression.Recipe Keywords:<ul><li>Medium: <code>amateur photo</code>, <code>smartphone photo</code>, <code>selfie</code>, <code>mirror selfie</code>.</li><li>Essential Styles: <code>visible sensor noise</code>, <code>jpeg artifacts</code>, <code>harsh shadows</code> (from sun), <code>blown-out highlights</code>, <code>direct flash</code> (for night shots).</li><li>AVOID: <code>motion blur</code>, <code>blockiness</code>, <code>video frame grab</code>.</li></ul>Example Prompt (Style 1):<code>1girl, amateur smartphone photo of an adult with a slender build for a casual street style portrait, posed leaning against a graffiti-covered wall on a sunny day, she faces the camera with a slight smile, her long blonde hair framing her face which has natural makeup and visible freckles, wearing a white ribbed crop top, low-rise baggy jeans, and white sneakers, with a thin chain necklace and multiple rings. The scene is defined by harsh daylight creating strong, harsh shadows on the ground and slightly blown-out highlights on her hair, with visible sensor noise in the darker areas of the brick wall, where graffiti subtly spells out "1GIRL".</code>STYLE 2: The Video Frame Grab LookThis style emulates a low-fidelity still frame grabbed from a video. It is defined by technical degradation from compression and low-light recording. It feels raw, unfiltered, and truly "in the moment."Recipe Keywords:<ul><li>Medium: <code>video frame grab</code>, <code>video still</code>, <code>older smartphone photo</code>, <code>lo-fi</code>.</li><li>Essential Styles: <code>significant grain</code>, <code>blocky jpeg artifacts</code>, <code>chroma noise</code>, <code>soft focus</code>, <code>crushed shadows</code>.</li><li>Use <code>motion blur</code> only if the pose is dynamic (e.g., dancing, turning quickly).</li></ul>Example Prompt (Style 2):<code>1girl, video frame grab of an adult with a curvy build dancing in a dimly lit, messy room at night, facing the camera with an open-mouthed, joyful expression. She wears a black crop top and plaid pajama pants, her face features glossy lips, and her skin texture is completely obscured by significant motion blur and heavy digital noise. The room is illuminated by the purple glow of a neon sign on the wall shaped like the word "1GIRL", creating a strong tungsten cast and crushed shadows. The entire image is a low-quality still, suffering from extreme motion blur, heavy grain, and blocky jpeg artifacts.</code><hr /><h3 id="the-anatomy-of-a-perfect-prompt">The Anatomy of a Perfect Prompt</h3>Structure your prompts like a forensic report. More detail is always better.<ol><li>Layer 1: Foundation. <code>1girl</code> + The Medium (<code>amateur photo</code>, <code>video frame grab</code>).</li><li>Layer 2: Narrative. The scene, pose, expression, and vibe. (e.g., "posed playfully in her messy room," "candidly walking down a city street at night").</li><li>Layer 3: Subject &amp; Attire. Be hyper-specific. "an adult with a slender build," "wearing a black lace-trimmed camisole, ripped denim shorts, and platform boots." Explicitly mention accessories or their absence (<code>no necklace</code>, <code>septum piercing</code>).</li><li>Layer 4: The Style Modifiers. This is where you command the LoRa's style. Combine 3-5 keywords from the list below to create authentic flaws.</li></ol><hr /><h3 id="the-style-modifiers:">The Style Modifiers:</h3>You must explicitly ask for imperfections to get the intended results. Mix and match from these categories.<ul><li>Lighting Flaws: Simulate uncontrolled, real-world lighting.<ul><li><code>direct flash</code>, <code>on-camera flash</code>, <code>harsh shadows</code>, <code>blown-out highlights</code>, <code>crushed shadows</code>, <code>mixed lighting</code>, <code>strong backlighting</code>, <code>lens flare</code>, <code>tungsten cast</code>, <code>neon light</code>.</li></ul></li><li>Digital Artifacts &amp; Quality: Mimic the limitations of phone sensors and file compression.<ul><li><code>visible sensor noise</code>, <code>grainy</code>, <code>chroma noise</code>, <code>jpeg artifacts</code>, <code>blockiness</code>, <code>color banding</code>, <code>motion blur</code>, <code>soft focus</code>, <code>low resolution</code>.</li></ul></li><li>Processing &amp; Filters: Replicate common social media effects.<ul><li><code>beauty filter</code>, <code>skin smoothing</code>, <code>plastic skin appearance</code>, <code>oily shine</code>.</li></ul></li></ul><hr /><h3 id="advanced-technique:-the-&quot;1girl&quot;-text-feature">Advanced Technique: The "1GIRL" Text Feature</h3>Qwen has excellent text rendering. We can use this to subtly brand our images. The key is to make the text "1GIRL" feel like a natural part of the environment, not a watermark.Creative Integration Ideas:<ul><li>A faded graffiti tag on a wall.</li><li>A glowing neon sign in a bar or room.</li><li>Text on a custom phone case visible in a mirror selfie.</li><li>A license plate in the background of a street shot (<code>CA 1GIRL</code>).</li><li>The name of a fictional band on a poster or t-shirt.</li><li>A sticker on a laptop, mirror, or skateboard.</li></ul><hr /><h3 id="pro-tips-for-a-tier-aesthetics">Pro-Tips for A-Tier Aesthetics</h3>To guide the model towards generating more compelling subjects and avoid a generic "mid" look:<ul><li>Direct the Face: Use detailed facial descriptions. <code>heavy makeup with winged eyeliner and glossy lips</code>, <code>natural makeup with visible freckles</code>, <code>pensive expression</code>, <code>slight smile with direct gaze</code>, <code>face obscured by hair</code>.</li><li>Be Specific with Attire: The LoRa is trained on specific fashion subcultures. Use keywords that evoke them. <code>y2k street style</code>, <code>goth, alt style</code>, <code>athleisure wear</code>. Instead of "shirt," say "oversized graphic tee." Instead of "pants," say "low-rise baggy cargo pants."</li><li>Command the Pose: Don't leave the pose to chance. <code>looking back over her shoulder</code>, <code>hands on her hips</code>, <code>leaning against a car</code>, <code>squatting for a mirror selfie</code>. This adds dynamism and confidence.</li></ul><hr /><h3 id="instara-fair-use-license">Instara Fair Use License</h3>Our license is simple: Use our models freely for your own creative and commercial projects. Do not re-sell our models or use them on a competing service.By downloading Instara models, you agree to the legally binding terms outlined below.<h3 id="1.-the-fair-use-grant">1. The Fair Use Grant</h3><ul><li>You CAN Generate Images for Any Purpose: You can use images created with this model for personal, artistic, and commercial projects.</li><li>You CAN Sell Your Generations: You can sell the images you create. This includes prints, commissions, and use in your own commercial products or on monetized platforms (Patreon, YouTube, etc.).</li><li>You CAN Merge and Mix: You are free to merge this model with other models.</li></ul><h3 id="2.-core-restrictions">2. Core Restrictions</h3><ul><li>You MAY NOT Sell This Model: You are strictly prohibited from selling this LoRA file, its derivatives, or any model merges that include it.</li><li>You MAY NOT Host This Model on Generation Services: You are strictly prohibited from hosting or using this model on any public or private, commercial or free, image/video generation service, platform, application, or API.</li></ul><h3 id="3.-general-terms-and-conditions">3. General Terms &amp; Conditions</h3><ul><li>Attribution: While not required for commercial use of generated images, crediting Instara when sharing is appreciated and helps our community grow.</li><li>Intellectual Property: All rights, title, and interest in and to the model files themselves remain the exclusive property of Instara. This license grants you rights to the output of the model, not ownership of the model itself.</li><li>Ethical Use: You may not use this model for any purpose that is unlawful, defamatory, harassing, or hateful. You may not use it to create content that impersonates real individuals or to generate deceptive or harmful misinformation.</li><li>Disclaimer of Warranty: The model is provided "AS IS," without warranty of any kind. Instara shall not be liable for any damages arising from the use of this model.</li><li>Termination: This license will terminate automatically if you fail to comply with any of its terms. Upon termination, you must cease all use of the model.</li></ul>For enterprise licensing or platform integration inquiries, please contact us at <code>legal@instara.io</code>

1GIRL QWEN-IMAGE

1GIRL v2.0 is a significant update focused on increasing model diversity and detail fidelity.<ul><li>Refined Dataset: This version is trained on a new, highly-refined dataset of about the same number of images, however we focused on targeting perfect anatomy and skin texture while upping the image quality as well.</li></ul>

Files with SHA256: 853cbdad433444e23024b3a60523845c243ed10e663279dbce0cb15bc7afec0f