D.R.E.A.M

<div align="center">

🌌 D.R.E.A.M.

Digital Rendering Engine for Artistic Melodies

Diffusion-based text-to-image and image-to-image generation, tuned for artistic, anime, and photorealistic rendering.

</div>

---

📖 Overview

D.R.E.A.M. is the original model in the D.R.E.A.M. family — a suite of diffusion-based text-to-image and image-to-image models built to emulate the creativity and visual richness of tools like Midjourney, with specialized variants for different artistic domains.

Turn a sentence into a scene. Turn a scene into a style.

---

🧩 Checkpoints in This Repo

Unlike D.r.e.a.m_Mega (a full Diffusers pipeline repo), this repo ships its variants as individual single-file checkpoints — load whichever one fits your style with from_single_file().

(Filenames like -Dsh-Cet-(0.5) and -DRL-PR-RV(0.3) indicate a merge and its blend weight — update this row with the exact source models and ratios if you want that documented.)

Every checkpoint above supports both text-to-image and image-to-image generation. Looking for a full Diffusers pipeline folder instead of a single-file checkpoint? Use Cyanex/D.r.e.a.m_Mega instead.

---

🗺️ How It Works

 📝 Text Prompt
       │
       ▼
 🔤 Text Encoder
       │
       ▼
 🌫️  Diffusion U-Net
       │
       ▼
   Model Variant?
   ┌───────┼────────────┐
   ▼       ▼            ▼
🎨 Base  🌸 Anime   📸 Photorealism
   │       │            │
   └───────┴────────────┘
              │
              ▼
       🖼️ Generated Image

---

✨ Features

• 🖋️ Text-to-Image Generation — Turn descriptive prompts into original images
• 🛠️ Image-to-Image — Guide edits to existing images with new prompts
• 🎭 Creative Rendering — Artistic, imaginative, Midjourney-inspired outputs
• 🧬 Family of Variants — Pick the model tuned for your visual style (base, Mega, Anime, Photorealism)

---

🚀 Usage

pip install -U diffusers transformers accelerate

Since these are single-file checkpoints, load them with from_single_file() and the direct file URL — pick whichever variant fits your prompt:

import torch
from diffusers import StableDiffusionPipeline

# Base / general-purpose
ckpt_url = "https://huggingface.co/Cyanex/D.R.E.A.M/blob/main/D.R.E.A.M-Forge.ckpt"

# Anime — swap ckpt_url for:
# "https://huggingface.co/Cyanex/D.R.E.A.M/blob/main/D.R.E.A.M_Anime-Dsh-Cet-(0.5).safetensors"

# Photorealism — swap ckpt_url for:
# "https://huggingface.co/Cyanex/D.R.E.A.M/blob/main/D.r.e.a.m_Photorealism-DRL-PR-RV(0.3).safetensors"

pipe = StableDiffusionPipeline.from_single_file(ckpt_url, torch_dtype=torch.float16)
pipe = pipe.to("cuda")

prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k"
image = pipe(prompt).images[0]
image.save("dream_output.png")

These same files also drop straight into AUTOMATIC1111, ComfyUI, or other webui-style tools that support .ckpt/.safetensors checkpoints — just download the file into your models/Stable-diffusion folder.

🛠️ Image-to-Image

Any of the checkpoints above can also restyle or modify an existing image — load with AutoPipelineForImage2Image instead:

import torch
from PIL import Image
from diffusers import AutoPipelineForImage2Image

ckpt_url = "https://huggingface.co/Cyanex/D.R.E.A.M/blob/main/D.R.E.A.M-Forge.ckpt"

pipe = AutoPipelineForImage2Image.from_single_file(ckpt_url, torch_dtype=torch.float16)
pipe = pipe.to("cuda")

init_image = Image.open("input.png").convert("RGB")
prompt = "turn it into a vibrant Midjourney-style painting, dramatic lighting"

image = pipe(prompt=prompt, image=init_image, strength=0.6, guidance_scale=7.5).images[0]
image.save("dream_img2img_output.png")

---

⚙️ Model Details

---

⚠️ Limitations

• May struggle with fine anatomical detail (hands, small objects, in-image text)
• Prompt adherence can vary by subject and style
• Not intended for generating photorealistic depictions of real, identifiable people

---

🤝 Contributing

Contributions are welcome! Feel free to open an issue or pull request if you'd like to help improve the D.R.E.A.M. family — whether that's data, training tips, or example prompts.

---

<div align="center">

Made with 🌙 by Cyanex1702

</div>