A multiplexed offline ASR system using VAD segmentation and shared-batch Whisper inference reduces p90 latency versus sequential baseline, especially under high concurrency.
We compare this against the baseline of sequential processing, where each audio is pro- cessed completely before starting the next
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.AS 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Scalable Offline ASR for Command-Style Dictation in Courtrooms
A multiplexed offline ASR system using VAD segmentation and shared-batch Whisper inference reduces p90 latency versus sequential baseline, especially under high concurrency.