---
library_name: mlx
license: mit
base_model: Picsart-AI-Research/MI-GAN
pipeline_tag: image-to-image
tags:
  - mlx
  - inpainting
  - object-removal
---

# mlx-community/MI-GAN-256-ffhq-fp16

MI-GAN mobile inpainting, 256 FFHQ (faces).

Converted to **Apple MLX** (`-fp16`) for Apple-Silicon inference via the
[`mlx-lama-swift`](https://github.com/xocialize/mlx-lama-swift) Swift package (the MLXEngine
`imageInpaint` ModelPackage / Forge **Erase** capability). From-scratch MLX-Swift architecture port
of [Picsart-AI-Research/MI-GAN](https://github.com/Picsart-AI-Research/MI-GAN); parity-locked vs the PyTorch oracle on the CPU stream
(LaMa predicted max_abs 3.2e-5 · MI-GAN 5.5e-5).

## Use

```swift
// Package.swift → .package(url: "https://github.com/xocialize/mlx-lama-swift", from: "0.1.0")
import LaMa   // or MIGAN
let inpainter = try LaMaInpainter.fromPretrained(weightsPath, dtype: .float16)
let filled: CGImage = inpainter(sourceCGImage, mask: maskCGImage)  // white mask = remove
```

Input: image + mask (white = region to remove). Output: filled image at source resolution.
Weights license: mit. Port code: MIT.