Craftly-Image-2.1

🏗️ Architecture Overview

• Core Framework: Latent Diffusion Transformer (DiT) with Flow-Matching Schedulers
• Text Conditioning Backbone: CraftlyVision-7B-Instruct multimodal encoder
• Latent Autoencoder: High-fidelity 16-channel Spatial VAE
• Native Spatial Support: Dynamic aspect ratio synthesis up to 2K resolution

🚀 Quickstart & Inference

import torch
from diffusers import DiffusionPipeline

# Load full Craftly-Image-2.1 pipeline
pipe = DiffusionPipeline.from_pretrained(
    "CraftlyrobotMushfiqur/image",
    torch_dtype=torch.bfloat16
)
pipe.to("cuda")

prompt = "A breathtaking futuristic cyberpunk laboratory with neon reflections, ultra high resolution 8K"
image = pipe(prompt, num_inference_steps=30, guidance_scale=4.5).images[0]
image.save("craftly_output.png")

⚖️ Citation & Attribution

@article{craftlyimage2026,
  title={Craftly-Image-2.1: Advanced Visual Synthesis via Deep Multimodal Conditioning},
  author={Md Mushfiqur Rahim and Craftly AI Research Team},
  year={2026},
  publisher={Hugging Face}
}