SDXLTurboModel<T>

Models & Types · AiDotNet.Diffusion.FastGeneration

SDXL Turbo model for real-time single-step high-resolution image generation.

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

For Beginners

SDXL Turbo combines the image quality of SDXL (one of the best open-source models) with near-instant generation. While regular SDXL needs 25-50 steps, SDXL Turbo generates comparable images in just 1 step. No guidance needed (scale=0).

How It Works

SDXL Turbo is the SDXL-based variant of Adversarial Diffusion Distillation (ADD), generating 512x512 images in 1-4 steps with SDXL-quality aesthetics. Uses the full SDXL U-Net with dual text encoder conditioning (CLIP ViT-L + OpenCLIP ViT-bigG).

Reference: Sauer et al., “Adversarial Diffusion Distillation”, 2023

Example

var options = new DiffusionModelOptions<float> { LatentChannels = 4, DefaultInferenceSteps = 1 };
var noise = Tensor<float>.CreateRandom(new[] { 1, 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 SDXLTurboModel<float>(options: options))
    .Build(trainX, trainY);
var generated = result.Predict(noise);

Constructors

Constructor Summary
SDXLTurboModel(NeuralNetworkArchitecture<T> architecture, DiffusionModelOptions<T> options, INoiseScheduler<T> scheduler, UNetNoisePredictor<T> predictor, StandardVAE<T> vae, IConditioningModule<T> conditioner, int? seed) Initializes a new SDXL Turbo 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()