⚡ P-Q-Image-2.1
Few-step LoRA adapters for Qwen-Image-2.1 — ComfyUI
6 steps (v0.2), 5 or 8 steps (v0.1) · Up to 6.3x faster · No CFG · Text-to-image and image editing
P-Q-Image-2.1 is a set of LoRA adapters that let
Qwen/Qwen-Image-2.1 generate
and edit images in only 6 steps (v0.2), or 5 or 8 steps (v0.1). The adapters load
on top of the base model, so the pipeline, text encoder, and VAE stay unchanged.
This repository is the ComfyUI export of the original repo
PrunaAI/P-Q-Image-2.1,
plus example workflows for text-to-image and image editing.
The adapters load with the native LoraLoaderModelOnly node on top of the
Qwen-Image-2.1 base checkpoint from
Comfy-Org/Qwen-Image-2.1.
No custom nodes are needed. Update ComfyUI to a version with native Qwen-Image-2.1 support.
Versions
| Version | Variant | File | Steps |
|---|---|---|---|
| v0.2 | 6-steps |
p_q_image_2.1_6_steps_v0.2_comfyui_bf16.safetensors |
6 |
| v0.2 | 6-steps-upsampled |
p_q_image_2.1_6_steps_upsampled_v0.2_comfyui_bf16.safetensors |
6 |
| v0.1 | 8-steps |
p_q_image_2.1_8_steps_v0.1_comfyui_bf16.safetensors |
8 |
| v0.1 | 5-steps |
p_q_image_2.1_5_steps_v0.1_comfyui_bf16.safetensors |
5 |
For prompting guides, benchmarking, and limitations, see the original repo
PrunaAI/P-Q-Image-2.1.
Setup
Place the files in the following folders:
ComfyUI/
└── models/
├── diffusion_models/
│ └── qwen_image_2.1_int8_convrot.safetensors (or qwen_image_2.1_bf16.safetensors)
├── loras/
│ └── p_q_image_2.1_6_steps_v0.2_comfyui_bf16.safetensors
├── text_encoders/
│ └── qwen3vl_8b_int8_convrot.safetensors
└── vae/
└── qwen_image_2.1_vae_bf16.safetensors
The base model, text encoder and VAE come from Comfy-Org/Qwen-Image-2.1. Use the base checkpoint, not the Turbo one.
Workflows
Drag a file from workflows/ into ComfyUI.
| Workflow | Adapter |
|---|---|
pruna_q_image_2.1_v0.2_6step_t2i.json |
v0.2 6-steps (switch lora_name to 6-steps-upsampled) |
pruna_q_image_2.1_v0.2_6step_image_edit.json |
v0.2 6-steps (switch lora_name to 6-steps-upsampled) |
pruna_q_image_2.1_v0.1_8step_t2i.json |
v0.1 8-steps |
pruna_q_image_2.1_v0.1_8step_image_edit.json |
v0.1 8-steps |
pruna_q_image_2.1_v0.1_5step_t2i.json |
v0.1 5-steps |
pruna_q_image_2.1_v0.1_5step_image_edit.json |
v0.1 5-steps |
Each workflow loads one adapter at strength 1.0 and samples with euler,
cfg 1, and the matching ManualSigmas schedule.
Sampling settings
Use the sigma schedule that matches the adapter. Do not mix v0.1 and v0.2 schedules.
| Adapter | ManualSigmas |
|---|---|
| v0.2, 6 steps | 1.0, 0.961054, 0.918949, 0.769302, 0.545554, 0.222292, 0 |
| v0.1, 8 steps | 1.0, 0.933333, 0.857143, 0.769231, 0.666667, 0.545455, 0.4, 0.222222, 0 |
| v0.1, 5 steps | 1.0, 0.94, 0.857143, 0.666667, 0.4, 0 |
- cfg 1, no negative prompt, sampler
euler, LoRA strength 1.0. - The adapters were trained at 1K resolution only.
License
The end distilled model is based on Qwen-Image-2.1.
P-Q-Image-2.1 is a derivative of Qwen-Image-2.1 and is distributed under the Qwen RESEARCH LICENSE AGREEMENT. Review its use restrictions before using or redistributing the adapters. See LICENSE and NOTICE.
Qwen is licensed under the Qwen RESEARCH LICENSE AGREEMENT, Copyright (c) 2026 Hangzhou Tongyi Laboratory Technology Co., Ltd. All Rights Reserved.
What's next?
- Use P-Q-Image-2.1 to generate and edit images with Qwen-Image-2.1 in a few steps.
- Compress your own models with Pruna and give us a ⭐️ for more efficiency!
- Want to use our optimized image models right away? Check P-Image-Ideogram and P-Image-Edit.