Researchers have developed LILAC, a novel neural speech codec designed to be idempotent, meaning it can decode and re-encode audio without altering the original token stream. This is a significant improvement over existing codecs, which typically modify a substantial portion of tokens during such processes. LILAC maintains competitive audio quality, achieving UTMOS scores comparable to state-of-the-art codecs while operating at a low bitrate of 0.75 kbit/s. AI
IMPACT Introduces a more robust token interface for speech generation and editing pipelines.
RANK_REASON Academic paper detailing a new technical approach to neural audio codecs. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
