StepVideoModel<T>

Models & Types · AiDotNet.Diffusion.Video

StepVideo text-to-video model with benchmark-leading quality.

Inheritance: DiffusionModelBase<T> → LatentDiffusionModelBase<T> → VideoDiffusionModelBase<T> → StepVideoModel<T>

For Beginners

Step Video is a ~10B parameter text-to-video model that competes with state-of-the-art proprietary models. It excels at generating diverse, high-quality video content with strong prompt adherence.

How It Works

References:

  • Paper: “StepVideo: Text-to-Video Generation with Scalable Diffusion Transformers” (StepFun, 2025)

StepVideo is a ~10B parameter T2V model from StepFun competing with HunyuanVideo on quality benchmarks. It uses a scalable DiT architecture with deep text understanding, handling complex prompts and quality motion reliably. The model is released as open-source with weights and code.

Technical specifications:

  • Architecture: Scalable DiT + Deep Text Understanding
  • Latent channels: 16
  • Default: 51 frames at 16 FPS
  • Supports I2V: Yes | T2V: Yes | V2V: No

Example

var options = new DiffusionModelOptions<float> { LatentChannels = 16, DefaultInferenceSteps = 50 };
var noise = Tensor<float>.CreateRandom(new[] { 1, 16, 51, 68, 124 });
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 StepVideoModel<float>(options: options))
    .Build(trainX, trainY);
var video = result.Predict(noise);

Constructors

Constructor Summary
StepVideoModel(NeuralNetworkArchitecture<T> architecture, DiffusionModelOptions<T> options, INoiseScheduler<T> scheduler, DiTNoisePredictor<T> predictor, TemporalVAE<T> temporalVAE, IConditioningModule<T> conditioner, int defaultNumFrames, int defaultFPS) Initializes a new instance of StepVideoModel with full customization support.

Properties

Property Summary
IConditioningModule<T> Conditioner { get; }
INoisePredictor<T> NoisePredictor { get; }
IVAEModel<T> TemporalVAE { get; }
IVAEModel<T> VAE { get; }
bool SupportsImageToVideo { get; }
bool SupportsTextToVideo { get; }
bool SupportsVideoToVideo { get; }
int LatentChannels { get; }

Methods

Method Summary
ModelMetadata<T> GetModelMetadata()