UniPASE extends the PASE framework with DeWavLM-Omni to convert degraded speech into high-fidelity, low-hallucination audio across sampling rates via phonetic enhancement, acoustic adaptation, and multi-rate vocoding.
Miipher-2: A universal speech restoration model for million- hour scale data restoration,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.AS 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
UniPASE: A Generative Model for Universal Speech Enhancement with High Fidelity and Low Hallucinations
UniPASE extends the PASE framework with DeWavLM-Omni to convert degraded speech into high-fidelity, low-hallucination audio across sampling rates via phonetic enhancement, acoustic adaptation, and multi-rate vocoding.