5 papers
The Anatomy of an ASR Hallucination
Hamees Sayed, Apoorv Singh, Kumar Aman +1
ASR systems sometimes produce fluent text that is unrelated to the speech they receive. We view these hallucinations as one possible consequence of a broader grounding failure, in…
DLLM-TTS: Block Discrete Diffusion Language Model for Text-to-Speech Synthesis
Wasim Madha, Nityanand Mathur, Hamees Sayed +4
Current text-to-speech systems face a trade-off: autoregres- sive codec language models produce highly intelligible speech but require large-scale models and training data and deco…
How Do Instructions Shape Speech? Cross-Attention Attribution for Style-Captioned Text-to-Speech
Nityanand Mathur, Hamees Sayed, Wasim Madha +4
Style-captioned text-to-speech systems use natural language to control voice characteristics, but how individual words influence acoustic output remains unclear. Understanding this…
Rewriting TTS Inference Economics: Lightning V2 on Tenstorrent Achieves 4x Lower Cost Than NVIDIA L40S
Ranjith M. S., Akshat Mandloi, Sudarshan Kamath
Text-to-Speech (TTS) models are significantly more numerically fragile than Large Language Models (LLMs) due to their continuous waveform generation and perceptual sensitivity to s…
SonoEdit: Null-Space Constrained Knowledge Editing for Pronunciation Correction in LLM-Based TTS
Ayush Pratap Singh, Harshit Singh, Nityanand Mathur +2
Neural text-to-speech (TTS) systems systematically mispronounce low-resource proper nouns, particularly non-English names, brands, and geographic locations, due to their underrepre…