Pruna-Qwen-Image-2.1
Unofficial mirror. This listing mirrors the original project for reference and convenience. It is not created, maintained, or endorsed by PrunaAI. Please credit the original authors and preserve all license, attribution, and modification notices.
Official repositories: PrunaAI/P-Q-Image-2.1 · Official ComfyUI pack and workflows
Base model: Qwen/Qwen-Image-2.1
Overview
These LoRA adapters speed up text-to-image generation and image editing with Qwen-Image-2.1. They are not standalone models; the Qwen-Image-2.1 base checkpoint is required.
v0.2 — 6-step adapters
Two alternatives are available. Load only one at a time.
6-steps: p_q_image_2.1_6_steps_v0.2_comfyui_bf16.safetensors
Faithful base-model behavior; especially useful for editing where the input should be preserved.
6-steps-upsampled: p_q_image_2.1_6_steps_upsampled_v0.2_comfyui_bf16.safetensors
Improved results from shorter prompts without a separate prompt-upsampling call.
The upsampled adapter may not reproduce exactly what the original prompt upsampler would produce.
v0.1 — existing versions
The 8-step and 5-step adapters remain available. The 8-step version favors quality; the 5-step version is faster but has lower visual quality. Each adapter has its own sampling schedule.
ComfyUI settings
Use an updated ComfyUI version with native Qwen-Image-2.1 support. Load the adapter with the built-in LoraLoaderModelOnly node on the Qwen-Image-2.1 base checkpoint, not the Turbo checkpoint.
For either v0.2 adapter, use Euler, CFG 1, and LoRA strength 1.0. Set ManualSigmas to:
1.0, 0.961054, 0.918949, 0.769302, 0.545554, 0.222292, 0
Do not use the v0.1 schedule with a v0.2 adapter. No negative prompt or custom nodes are required.
Prompts and resolution
Start at 1024 × 1024. The adapters were trained at 1K resolution for text-to-image and editing, with up to three reference images. Higher resolutions, including 2K, and additional reference images are outside the training coverage, so results may vary. Detailed prompts are still useful; for the upsampled adapter, start with your original prompt.
License and attribution
This is a derivative of Qwen-Image-2.1 and is subject to the Qwen Research License Agreement. Review the license before use or redistribution. This mirror grants no additional rights; preserve the original attribution and license notices.
Description
FAQ
Comments (8)
Can you please put some more sample images, maybe examples of all types, t2i, i2i, base vs lora, etc. Thanks.
I’ve already uploaded some I2I and T2I examples, including comparisons between the base model and the LoRA, as you requested
@zoefollonier thank you
These Loras can't work with NSFW Loras.
能不能给个工作流
这里有一个适用于 Pruna Qwen Image 2.1 8-step LoRA 的 ComfyUI workflow，包含所需的自定义 sigma schedule、SamplerCustom 和 LoRA strength 1.0： https://civitai.red/models/2969889/pruna-8-steps-qwen-image-21-comfyui-workflow?modelVersionId=3365383 希望对你有帮助！
You can use workflow from my post - Image post by xNzX_IL | Civitai
1) Click on image (open page with one image)
2) Right click on image --> choose 'Save image as...'
3) Load this PNG image in your ComfyUI to get workflow.
For me, 5 steps and 8 steps LoRA gives better quality, then without LoRA. May be it result of great samplers res_2s from RES4LYF
https://github.com/ClownsharkBatwing/RES4LYF
Also, here is new LoRA with 6 steps:
https://huggingface.co/Viggle/Qwen-Image-2.1-viggle-turbo