Pruna-Qwen-Image-2.1 provides LoRA adapters for Qwen-Image-2.1, enabling text-to-image and image editing in 5 or 8 steps. The 8-step version is recommended for better quality, while the 5-step offers up to 6.3x faster generation without CFG scale.
The adapters operate at 1K resolution with up to 3 reference images for editing. They require the base Qwen-Image-2.1 model and are licensed under the Qwen RESEARCH LICENSE AGREEMENT.
As a work in progress (v0.1), the adapters do not match the base model's visual quality. They are designed for specific step counts and scheduler settings only.
Highlights
- Use the sigma schedule that matches your adapter, and keep the scheduler
- 8-step: 1 → 14/15 → 6/7 → 10/13 → 2/3 → 6/11 → 0.4 → 2/9 → 0
- 5-step: 1 → 0.94 → 6/7 → 2/3 → 0.4 → 0
- No CFG. Keep truecfgscale=1.0 and do not pass a negative prompt.
- Keep the LoRA strength at 1.0.
- Write detailed prompts for text-to-image. Longer prompts that describe
- This is a first version. Quality is below that of the base model.
- The 5-step adapter is faster than the 8-step one, but its images are
From the source
Images from PrunaAI on Hugging Face, shown with a link to the original.