mushfiqur-good-boyyy/m2-craftly-1

M2-Craftly (by Mushfiqur)

A private, rebranded build of aMUSEd, a lightweight text-to-image model using a MUSE-style (masked image modeling) architecture — U-ViT + VQ-GAN, ~800M parameters.

This is a rebrand, not a new architecture. Weights and architecture are unchanged from the upstream aMUSEd model; only the name, packaging, and model card are customized here for personal/private use.

• —Original model: amused/amused-256 (openrail++ license)
• —Rebrand: M2-Craftly by Mushfiqur
• —Visibility: Private (shared only with invited collaborators)

## Usage

    import torch
    from diffusers import AmusedPipeline as M2CraftlyPipeline  # same class, renamed on load

    pipe = M2CraftlyPipeline.from_pretrained("mushfiqur-good-boyyy/m2-craftly-1", torch_dtype=torch.float32)
    image = pipe("a cat on a rooftop at sunset", num_inference_steps=12).images[0]
    image.save("output.png")