FlowEdit stores pronunciation corrections for flow-matching TTS as token perturbations in a Modern Hopfield Network, cutting target-word phoneme error rate by 92.7% on a 312-word multilingual benchmark while preserving general speech quality.
Sonoedit: Null-space constrained knowledge editing for pronunciation correction in llm-based tts,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
FlowEdit: Associative Memory for Lifelong Pronunciation Adaptation in Flow-Matching TTS
FlowEdit stores pronunciation corrections for flow-matching TTS as token perturbations in a Modern Hopfield Network, cutting target-word phoneme error rate by 92.7% on a 312-word multilingual benchmark while preserving general speech quality.