LavaSR

LavaSR is a lightweight and high-quality speech enhancement model designed to improve low-quality audio inputs. It utilizes a 50MB bandwidth extension model combined with the UL-UNAS denoiser to restore and enhance speech effectively.

LavaSR is a lightweight and high-quality speech enhancement model designed to improve low-quality audio inputs. It utilizes a 50MB bandwidth extension model combined with the UL-UNAS denoiser to restore and enhance speech effectively.

• 0150MB model size for efficient deployment
• 02Supports input rates from 8-48kHz, outputting at 48kHz
• 03Inference speeds ranging from 10-50x real-time on CPU and 400-4000x real-time on GPU
• 04Enhances low-quality audio datasets, TTS, ASR models, and voice calls