Turns any scene into a real-looking Polaroid instant photo: white frame with the wide bottom border, faded colors, soft focus, hard on-camera flash and that slightly imperfect chemical look. The Qwen version is punchier and more contrasty than the MOSS one, closer to a flash snapshot.Works on portraits, nightlife flash shots, travel, cars, still life. It doesn't care about the subject, only the look.HOW TO USETrigger word: p0laroidStart your prompt like this:p0laroid, a polaroid instant photo, &lt;your scene&gt;, white polaroid border with a wider bottom edgeRecommended settings• LoRA weight: 0.7 (0.6–0.8). Above 0.8 the image gets too soft and backgrounds turn to mush• Size: 896×1088 (true Polaroid proportions with the frame)• 25 steps, CFG 1, euler / simple (official ComfyUI template settings)• Keep backgrounds simple: walls, sky, curtains, cars. Very busy detail (frescoes, chandeliers) gets smeared• Avoid the words "frame" / "cropped" in the scene description. The model reads them as "camera frame" and drops the border.All preview images were made locally in ComfyUI with the official Qwen Image 2.1 int8 template + this LoRA at 0.7. Prompts are in the image metadata.Also available for MOSS Image. Look for "Polaroid Instant Film | MOSS Image" on my profile.<h3 id="more-workflows-loras-and-guides:-seedraai.com">More workflows, LoRAs and guides: <a target="_blank" rel="ugc" href="http://seedraai.com">seedraai.com</a></h3>DATASET &amp; CREDITSTrained on 54 real Polaroid photos from Wikimedia Commons, all under free licenses (CC BY / CC BY-SA). Huge thanks to the photographers:Sasha Kargaltsev, Makia Minich, Johan Hansson, C. J. Lee, Kevin Dooley, Tom Hart, Moki8, <a target="_blank" rel="ugc" href="http://Yellow.Cat">Yellow.Cat</a>, Peter Pawlowski, Tony Webster, Sgroey, Academicien, Mrnerd1billion, Nick McManus, Minimalist100.Licenses: CC BY 2.0, CC BY 3.0, CC BY 4.0, CC BY-SA 3.0, CC BY-SA 4.0.=== FORM FIELDS ===Type: LoRACategory: styleTags: polaroid, instant film, analog, film photography, vintage, flash photography, photorealistic, stylePermissions: same as MOSS

SEEDRAAI | Polaroid Instant Film | Qwen Image 2.1

First release. Trained on Qwen Image 2.1, 54 real Polaroid photos, 3000 steps, rank 32. Best at weight 0.7, 896×1088, 25 steps, CFG 1, euler / simple.

p0laroid_qwen_v1.safetensors

v1.0


# 

## Model Introduction

This model is trained using the [Civision/model training](https://modelscope.ai/civision/modelTraining) environment and computing resources provided by ModelScope.

* Model Type: LoRA
* Base Model: [Qwen/Qwen-Image-2.1](https://modelscope.ai/models/Qwen/Qwen-Image-2.1)
* Training Code: [DiffSynth-Studio](https://github.com/modelscope/DiffSynth-Studio)
* Data amount：54
* Total training steps：3000
* License: Apache-2.0

## Inference Code

Install [DiffSynth-Studio](https://github.com/modelscope/DiffSynth-Studio):

```bash
pip install diffsynth
```

Inference:

```python
from diffsynth.pipelines.qwen_image_21 import QwenImage21Pipeline, ModelConfig
import torch

pipe = QwenImage21Pipeline.from_pretrained(
    torch_dtype=torch.bfloat16,
    device="cuda",
    model_configs=[
        ModelConfig(model_id="Qwen/Qwen-Image-2.1", origin_file_pattern="transformer/diffusion_pytorch_model*.safetensors"),
        ModelConfig(model_id="Qwen/Qwen-Image-2.1", origin_file_pattern="text_encoder/model*.safetensors"),
        ModelConfig(model_id="Qwen/Qwen-Image-2.1", origin_file_pattern="vae/diffusion_pytorch_model*.safetensors"),
    ],
    processor_config=ModelConfig(model_id="Qwen/Qwen-Image-2.1", origin_file_pattern="processor/"),
)
pipe.load_lora(pipe.dit, ModelConfig(model_id="cfgscale/p0laroid_qwen_v1", origin_file_pattern="p0laroid_qwen_v1_c1-st3000.safetensors"))

prompt = "a cat"
image = pipe(prompt, seed=0, num_inference_steps=50)
image.save("image.png")
```


Files with SHA256: d0832a8ecd06b6d3c4292427e6901c0e7684720518e3a4774267e84a8a39d49c