Convtasnet Libri3Mix Sepnoisy 16k
ConvTasNet model trained based on the Asteroid framework for separating noise and speech signals from mixed audio.
Downloads 473
Release Time : 3/2/2022
Model Overview
This model was trained on the Libri3Mix dataset for noise separation tasks, capable of extracting clear speech signals from mixed audio containing multiple speakers and background noise.
Model Features
Efficient audio separation
Utilizes the ConvTasNet architecture to efficiently separate speech signals from multiple speakers in mixed audio.
Noise suppression
Specifically trained on mixed audio with background noise, effectively suppressing noise.
High sampling rate support
Supports 16kHz audio sampling rate, suitable for high-quality audio processing requirements.
Model Capabilities
Audio separation
Noise suppression
Speaker separation
Use Cases
Speech enhancement
Conference recording enhancement
Separates clear speech signals from conference recordings containing multiple speakers and background noise.
SI-SDR improvement of 10.28dB, SIR improvement of 18.57dB
Audio post-processing
Film and television audio processing
Separates clear dialogue signals from film and television recordings, removing background noise.
STOI improvement of 0.207, significant enhancement in speech clarity
Featured Recommended AI Models
Š 2025AIbase