<pre><code># IDK_SDXL-depth-of-field-camera — Depth of Field LoRA for SDXL

## Description

Adds realistic shallow depth of field, cinematic bokeh, and background blur. Perfect for portrait photography style and dramatic compositions.

---

### Before we start...

My therapist said I need to focus on what's important and blur out the rest.
So I trained a LoRA for that.
Turns out AI has better focus than I do.

&gt; **Disclaimer:** The dark humor above is just a joke to break the ice and does NOT reflect my actual feelings or beliefs. Just vibes. Enjoy the LoRA!

---

## Trigger Words

```
shallow depth of field, heavily blurred background, heavy bokeh, cinematic bokeh, depth of field
```

## Recommended Settings

| Setting | Recommended |
|---|---|
| **Base Model** | Any SDXL checkpoint |
| **Sampler** | Euler a, DPM++ SDE Karras, DPM++ 2M Karras |
| **Steps** | 25-35 |
| **CFG Scale** | 5-8 |
| **LoRA Weight** | 0.6 - 0.85 (start at 0.7, adjust to taste) |
| **Resolution** | 1024x1024, 896x1152, 832x1216 |
| **VAE** | SDXL default / sdxl-vae-fp16-fix |

## Tips

- Start with LoRA weight around **0.7** and adjust up/down depending on how strong you want the effect.
- Combine with other IDK_ LoRAs from this series for layered results (e.g. body + hair + style).
- Works with both realistic and stylized SDXL checkpoints.
- Use the trigger words in your prompt for best activation.
- If the effect is too subtle, increase weight or repeat the trigger word.
- If the effect is too strong / distorted, lower the weight to 0.5-0.6.

## Part of the IDK Series

This LoRA is part of the **IDK LoRA collection** — a set of specialized SDXL LoRAs that can be used individually or merged together. The merged version (**IDK.safetensors**) combines all of them using hierarchical semantic merging for maximum quality.

---

*Trained on SDXL 1.0 | LoRA Rank 16 | FP16*

---

## Prompt for Batch Generation (infinite mode)

Prompts use to generate the image of the main post, useing dynamic wildcards `{A|B}` for automatic ramdowness in infinite mode for the SD.NEXT.

**Positive:**
```
{portrait of a woman|a cat sitting on a fence|a flower in a garden|a person reading a book in a cafe|a musician playing guitar on stage}, shallow depth of field, {heavy bokeh|cinematic bokeh|creamy bokeh}, heavily blurred background, {city lights bokeh behind|nature bokeh|neon lights blurred|warm interior bokeh}, sharp subject focus, {85mm lens|135mm lens|50mm f1.4}, professional photography, masterpiece, best quality
```

**Negative:**
```
(extra fingers, deformed iris, blurry, lowres, bad anatomy, jpeg artifacts, compression artifacts, noise, grainy, motion blur), (low quality, worst quality:1.2), (compression artifacts, jpeg artifacts:1.4), (blurry, out of focus:1.2), grainy, noise, text, watermark, signature, error, messy lines
```
</code></pre><p></p>

Files with SHA256: bf68101881011cfa125f5212c2b59f98dd5f706a3b7f135c79aa08a36167c54b