orpatashnik/styleclip 🖼️📝🔢 → 🖼️

▶️ 1.3M runs 📅 May 2021 ⚙️ Cog 0.4.4 🔗 GitHub 📄 Paper ⚖️ License

clip image-editing image-manipulation latent-space stylegan text-driven text-to-image

About

Text-Driven Manipulation of StyleGAN Imagery

Example Output

Output

[object Object]

Performance Metrics

2.92s Total Time

All Input Parameters

{
  "input": "https://replicate.delivery/mgxm/d7df95aa-ffeb-4754-ba5e-b76b2eea5238/67.jpg",
  "target": "a face with a bowlcut",
  "neutral": "a face",
  "manipulation_strength": "4.1",
  "disentanglement_threshold": "0.15"
}

Input Parameters

input (required) Type: string: Input image
target Type: stringDefault: a face with a bowlcut: Target image description
neutral Type: stringDefault: a face: Neutral image description
manipulation_strength Type: numberDefault: 4.1Range: -10 - 10: The higher the manipulation strength, the closer the generated image becomes to the target description. Negative values moves the generated image further from the target description
disentanglement_threshold Type: numberDefault: 0.15Range: 0.08 - 0.3: The higher the disentanglement threshold, the more specific the changes are to the target attribute. Lower values mean that broader changes are made to the input image

Output Schema

Output

Type: string • Format: uri

Version Details

Version ID: 7af9a66f36f97fee2fece7dcc927551a951f0022cbdd23747b9212f23fc17021
Version Created: October 14, 2022

Run on Replicate →