StableVideoSROptions

Options & Configuration · AiDotNet.Video.Options

Configuration options for the StableVideoSR temporal-conditioned diffusion video super-resolution model.

Inheritance: ModelOptions → NeuralNetworkOptions → StableVideoSROptions

For Beginners

StableVideoSR takes the popular Stable Diffusion image AI and extends it to handle video. It adds special “temporal” modules that look at neighboring frames to ensure the output video is smooth and flicker-free, not just a sequence of independently upscaled images.

How It Works

Options for the Upscale-A-Video compatibility entry point. The implementation uses a temporal U-Net, temporal VAE decoder, low-resolution noise conditioning, overlapping temporal windows, DDIM sampling, and flow-guided recurrent latent propagation.

Constructors

Constructor Summary
StableVideoSROptions() Initializes a new instance with default values.
StableVideoSROptions(StableVideoSROptions other) Initializes a new instance by copying from another instance.

Properties

Property Summary
OnnxModelOptions OnnxOptions { get; set; } Gets or sets the ONNX runtime options.
VideoModelVariant Variant { get; set; } Gets or sets the model variant.
bool EnableFlowGuidedPropagation { get; set; } Enables the paper’s bidirectional flow-guided recurrent latent propagation.
double DropoutRate { get; set; } Gets or sets the dropout rate.
double GuidanceScale { get; set; } Gets or sets the classifier-free guidance scale.
double LatentScaleFactor { get; set; } Stable Diffusion x4-upscaler VAE latent scale.
double LearningRate { get; set; } Gets or sets the learning rate.
int LatentDim { get; set; } Gets or sets the latent space dimension.
int MaximumNoiseLevel { get; set; } Maximum low-resolution noise conditioning level.
int NoiseLevel { get; set; } Low-resolution degradation noise level passed to the 1,000-entry class embedding.
int NumDenoisingSteps { get; set; } Gets or sets the number of denoising steps.
int NumFeatures { get; set; } Gets or sets the number of UNet feature channels.
int NumTemporalLayers { get; set; } Compatibility alias for the pre-fidelity option name.
int NumTemporalModules { get; set; } Gets or sets the released count of stage-level temporal modules.
int ScaleFactor { get; set; } Gets or sets the spatial upscaling factor.
int TemporalWindowOverlap { get; set; } Number of frames blended between adjacent temporal windows.
int TemporalWindowSize { get; set; } Number of frames in each overlapping temporal denoising window.
int WarmupSteps { get; set; } Gets or sets the warmup steps.
int[] PropagationSteps { get; set; } Zero-based denoising iteration indices at which flow-guided x0 propagation runs.
string ModelPath { get; set; } Gets or sets the path to the ONNX model file.
string NegativePrompt { get; set; } Released command-line negative prompt used for classifier-free guidance.
string Prompt { get; set; } Default text guidance prompt.

Methods

Method Summary
void ValidateNativePaperContract() Rejects settings that the native paper graph cannot honor.