# Dreamlike Photoreal 2.0

Dreamlike Photoreal 2.0 is a photorealistic text-to-image model based on Stable Diffusion 1.5.

**Warning:** This model may generate suggestive content. To avoid NSFW results, add `"nude, naked"` to the negative prompt.

---

### Usage Tips

- Adding the word **photo** to your prompt often improves photorealism.
- Non‑square aspect ratios work better for certain subjects:
  - For portraits, try vertical ratios (e.g., 640×896, 768×1024).
  - For landscapes, try horizontal ratios (e.g., 896×640, 1024×768).
- The model was trained on 768×768 images, so start with resolutions around that size (e.g., 768×768, 768×1024, 1024×768) for best results.

---

### Example Outputs

*(Preview images are available in the model repository.)*

---

### 🧨 Diffusers Usage

You can use this model with the `StableDiffusionPipeline` from the `diffusers` library:

```python
from diffusers import StableDiffusionPipeline
import torch

model_id = "dreamlike-art/dreamlike-photoreal-2.0"
pipe = StableDiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16)
pipe = pipe.to("cuda")

prompt = "photo, a church in the middle of a field of crops, bright cinematic lighting, gopro, fisheye lens"
image = pipe(prompt).images[0]

image.save("./result.jpg")
```

---

### License

This model is released under a **modified** CreativeML OpenRAIL‑M license.

- **Use restrictions:**
  - You are **not allowed** to host, fine‑tune, or run inference with the model or its derivatives on websites, apps, or similar services.
  - You are free to host the model card and files (without performing inference or fine‑tuning) on both commercial and non‑commercial platforms, provided you:
    - State the full model name (Dreamlike Photoreal 2.0).
    - Include a copy of this license.
    - Reference the official model card.
  - You are free to use the **outputs** (images) for commercial purposes, but only for teams of **10 or fewer** people.
- **Prohibited uses:** You may not use the model to deliberately produce or share illegal, harmful, or otherwise unethical content.
- The authors claim no rights over the images you generate; you are responsible for your use of the outputs, which must comply with the license terms.
- If you redistribute the model weights, you must include the same restrictions and provide a copy of the modified OpenRAIL‑M license to all recipients.