반응형
[Paper 리뷰] Speaking Clearly: A Simplified Whisper-based Codec for Low-Bitrate Speech Coding
Speaking Clearly: A Simplified Whisper-based Codec for Low-Bitrate Speech CodingSpeech codec은 acoustic fidelity와 semantic preservation 간의 inherent confilct가 존재함SimWhisper-CodecSemantically-capable model을 high-fidelity acoustic reconstruction에 대해 adapt특히 frozen, simplified Whisper encoder를 활용하여 external supervision 없이 semantic, acoustic preservation을 balancing논문 (ICASSP 2026) : Paper Link1. Intro..
Paper/Neural Codec
2026. 3. 26. 13:15
반응형