OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Image & Video AI

Adain Style

Adain Style is a Lua/Torch implementation of arbitrary image style transfer using Adaptive Instance Normalization, accompanying a research paper. It transforms content images with user-provided style images and provides controls for stylization strength, color preservation, mixing styles and applying different styles to masked image regions.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Adain Style — OpenViST project profile in Image & Video AI.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when you want a research-code workflow for transforming images with arbitrary reference styles and need controls such as stylization strength, color preservation, style interpolation or masked spatial control.

Who it is for

  • Researchers and developers exploring neural style transfer in Lua/Torch.
  • Image creators who want to stylize images with arbitrary reference styles.
  • Developers adapting image stylization or video stylization workflows.

Documented uses

  • Apply a chosen style image to a content image.
  • Adjust the degree of stylization or preserve the content image’s colors.
  • Blend multiple style images using interpolation weights.
  • Apply different styles to foreground and background regions with a binary mask.
  • Process video through the documented video-stylization scripts.
Before you choose

Requirements & limits

The public GitHub repository identifies the code as MIT-licensed. The documented implementation uses Lua/Torch and requires Torch7; no hosted service is described.

  • This is repository code, not a hosted image-editing service; the documented workflow runs scripts in a Torch environment.
  • The model weights are obtained separately using models/download_models.sh.
  • CUDA, cuDNN and related Torch packages are optional; ffmpeg is listed for video processing.
  • The README’s speed figure is specific to 512 × 512 images on a Pascal Titan X.
  • The README documents image and video use, but video processing follows a separate script workflow.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Apply a style to one content image

Generate a stylized image from a content image and a style image using the documented basic command.

Before you start

  • Lua/Torch environment with Torch7, as required by the repository.
  • A pretrained decoder and modified VGG-19 network downloaded using the repository’s model script.
  • Paths to a content image and a style image.

Steps

  1. Download the model files with the command below.

    Documented command
    bash models/download_models.sh

  2. Run th test.lua -content PATH_TO_CONTENT_IMAGE -style PATH_TO_STYLE_IMAGE, substituting the paths to the two images.

  3. Optionally add documented controls such as -alpha to adjust stylization strength, -preserveColor to retain content colors, or -crop to center-crop both images.

Expected outcome

The test script processes the specified content and style images and writes a stylized result to the output directory.

Guide limitations

  • CUDA, cuDNN and related Torch packages are optional rather than required by the listed dependencies.
  • The repository reports 15 FPS for 512 × 512 images on a Pascal Titan X; that figure does not establish performance on other systems.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
16 October 2026
Original website title
xunhuang1995/AdaIN-style · Official homepage title · retained in its source language
How we review sourcesSuggest a correction