LLaSE-G1 A FOSS speech enhancement model (huggingface.co)

LLaSE-G1 is a unified speech enhancement model capable of handling multiple tasks without extra task prompts, including:

    Noise Suppression (SE)
    Target Speaker Extraction (TSE)
    Packet Loss Concealment (PLC)
    Acoustic Echo Cancellation (AEC)
    Speech Separation (SS)

huggingface link:

> https://huggingface.co/ASLP-lab/LLaSE-G1

demo page:

> https://submission-papers.github.io/LLaSE-G1-demo-page/

arxiv paper:

> https://arxiv.org/abs/2503.00493