One LoRA. More Creative Freedom.
Generate images. Edit existing images. Explore more creative possibilities — all with a single LoRA.
Introducing Qwen-Image-2.1 Uncensored Omni, an experimental rank-128 LoRA developed by Chatr Labs, fine-tuned from Qwen-Image-2.1.
Designed for greater creative flexibility, diverse poses, complex compositions, and unrestricted experimentation within applicable usage policies.
Whether you're creating original characters, exploring challenging compositions, or editing existing images, Omni brings both workflows together in one adapter.
✨ Key Features
🎨 Text-to-Image Generation
Create images directly from natural-language prompts using the Qwen-Image-2.1 base model.
🖼️ Image Editing
Transform existing images with natural-language instructions. Change outfits, environments, colors, and other visual elements using the same LoRA.
🧍 Pose Freedom
Designed to support diverse character poses, complex body positioning, and challenging compositions.
🔓 Uncensored Inference
The provided inference code does not add a prompt filter or output safety checker. Actual results depend on the base model, prompts, and deployment environment.
⚡ One LoRA, Two Workflows
Use the same adapter for both text-to-image generation and image editing.
⚙️ Recommended Settings
SettingRecommendedBase ModelQwen-Image-2.1LoRA Strength1.0Sampling Steps25True CFG4.0Negative Prompt" " (single space)Resolution1024 × 1024KV CacheEnabled
Important: Use a single space as the negative prompt when using the tested Diffusers pipeline. An empty negative prompt disables true CFG in that implementation.
This is a standard LoRA, not a few-step distilled model.
🚀 Getting Started
1. Download the LoRA
Download the .safetensors file from this page.
2. Download the Base Model
Qwen-Image-2.1 on Hugging Face
3. Load the LoRA
Load the adapter on top of Qwen-Image-2.1 with a compatible inference pipeline.
4. Start Creating
Use the recommended settings above as a starting point.
The LoRA is specifically trained for Qwen-Image-2.1 and is not compatible with earlier Qwen-Image architectures.
For tested Diffusers code, installation requirements, and detailed instructions, see our Hugging Face repository.
🧪 Version 0.1 — Experimental Release
This is our first public experimental release.
We're actively collecting community feedback to guide future improvements.
We'd love to hear:
What works particularly well?
Which poses or compositions are still challenging?
What should we improve in v0.2?
What additional workflows would you like to see?
Share your creations, report issues, and help us improve Omni.
🙏 Acknowledgements
This release was developed and evaluated with GPU infrastructure support from [Chatr AI]
Special thanks to the Chatr AI team for providing the compute resources that helped make Omni v0.1 possible.
🔗 Links
Hugging Face — Model Weights & Documentation
https://huggingface.co/chatrlabs/Qwen-Image-2.1-Uncensored-Omni-LoRA-v0.1
GitHub — Chatr Labs
Official Website
Base Model
https://huggingface.co/Qwen/Qwen-Image-2.1
📜 License & Attribution
This adapter is based on Qwen-Image-2.1 and is distributed subject to the applicable Qwen Research License Agreement.
Use is limited to non-commercial research and evaluation unless separately authorized by the upstream rights holder.
See the Hugging Face repository for the license, attribution, and technical documentation.
Built with Qwen.
An independent community fine-tune by Chatr Labs. Not affiliated with or endorsed by the Qwen team.
Description
V0.1
FAQ
Comments (3)
Did you look at your example photos before you posted them? Lol.
haha what about it
@chatrlabs The eyes are... idk theres something that looks off