StyleAlignedModel<T>

Models & Types · AiDotNet.Diffusion.Control

Style-Aligned model for consistent style across multiple generated images.

Inheritance: DiffusionModelBase<T> → LatentDiffusionModelBase<T> → StyleAlignedModel<T>

For Beginners

When you generate multiple images, this model makes them all look like they belong to the same “art collection” — same colors, same style, same artistic feel. It’s like having one artist draw several different scenes.

How It Works

Style-Aligned uses shared self-attention across multiple generated images during the denoising process to ensure consistent style. This enables generating sets of images that share the same artistic style without explicit style transfer.

Reference: Hertz et al., “Style Aligned Image Generation via Shared Attention”, CVPR 2024

Example

// Create Style-Aligned model for consistent style across batches
var options = new DiffusionModelOptions<float>
{
    LatentChannels = 4,
    DefaultInferenceSteps = 30
};
            
// Generate multiple images with consistent artistic style
var noise = Tensor<float>.CreateRandom(new[] { 4, 4, 64, 64 });
var trainX = Tensor<float>.CreateRandom(4, 8);
var trainY = Tensor<float>.CreateRandom(4, 2);
var result = new AiModelBuilder<float, Tensor<float>, Tensor<float>>()
    .ConfigureModel(new StyleAlignedModel<float>(options: options))
    .Build(trainX, trainY);
var styledBatch = result.Predict(noise);

Constructors

Constructor Summary
StyleAlignedModel(NeuralNetworkArchitecture<T> architecture, DiffusionModelOptions<T> options, INoiseScheduler<T> scheduler, UNetNoisePredictor<T> baseUNet, StandardVAE<T> vae, IConditioningModule<T> conditioner, double styleAlignmentStrength, int? seed) Initializes a new Style-Aligned model.

Properties

Property Summary
IConditioningModule<T> Conditioner { get; }
INoisePredictor<T> NoisePredictor { get; }
IVAEModel<T> VAE { get; }
int LatentChannels { get; }

Methods

Method Summary
ModelMetadata<T> GetModelMetadata()