ACEStep<T>

Models & Types · AiDotNet.Audio.Generation

ACE-Step accelerated consistency-enhanced music generation model.

Inheritance: NeuralNetworkBase<T> → AudioNeuralNetworkBase<T> → ACEStep<T>

Implements: IAudioGenerator<T>

For Beginners

ACE-Step generates music from text descriptions super fast. While most AI music generators need many steps (like painting layer by layer), ACE-Step can create music in just 1-4 steps, making it fast enough for real-time use. You describe the music you want and it creates it almost instantly.

Usage:

How It Works

ACE-Step (2024) uses consistency training to generate high-quality music in just 1-4 diffusion steps instead of the 50-100 steps needed by standard models. It achieves real-time music generation while maintaining quality comparable to multi-step models, making it practical for interactive applications.

Constructors

Constructor Summary
ACEStep(NeuralNetworkArchitecture<T> architecture, ACEStepOptions options, IGradientBasedOptimizer<T, Tensor<T>, Tensor<T>> optimizer) Creates an ACE-Step model in native training mode.
ACEStep(NeuralNetworkArchitecture<T> architecture, string modelPath, ACEStepOptions options) Creates an ACE-Step model in native training mode.

Properties

Property Summary
bool IsOnnxMode { get; }
bool SupportsAudioContinuation { get; }
bool SupportsAudioInpainting { get; }
bool SupportsTextToAudio { get; }
bool SupportsTextToMusic { get; }
double MaxDurationSeconds { get; }

Methods

Method Summary
AudioGenerationOptions<T> GetDefaultOptions()
ModelMetadata<T> GetModelMetadata()
Task<Tensor<T>> GenerateAudioAsync(string prompt, string negativePrompt, double durationSeconds, int numInferenceSteps, double guidanceScale, int? seed, CancellationToken cancellationToken)
Tensor<T> ContinueAudio(Tensor<T> inputAudio, string prompt, double extensionSeconds, int numInferenceSteps, int? seed)
Tensor<T> GenerateAudio(string prompt, string negativePrompt, double durationSeconds, int numInferenceSteps, double guidanceScale, int? seed)
Tensor<T> GenerateMusic(string prompt, string negativePrompt, double durationSeconds, int numInferenceSteps, double guidanceScale, int? seed)
Tensor<T> InpaintAudio(Tensor<T> audio, Tensor<T> mask, string prompt, int numInferenceSteps, int? seed)