Skip to navigation

Hindi

Pulse streaming Hindi accuracy across public datasets.
View as Markdown

Hindi - multi-dataset (Streaming)

WER across seven Hindi speech datasets covering read speech (FLEURS), conversational speech (Kathbath, Common Voice), telephony / contact-center audio (Mucs, Gramvaani), TTS-derived audio (Indic-TTS), and a noise-augmented variant (Kathbath noisy). Compared against the strongest Hindi STT baselines: IndicWhisper, Sarvam Saaras v3, scribe v2, and Deepgram Nova-3.

Evaluated on the open-source datasets. Smallest Pulse numbers from internal evaluation. Lower is better.

DatasetSmallest PulseIndicWhisperSarvam Saaras v3scribe v2Deepgram Nova-3
FLEURS6.1715.006.188.9614.09
Kathbath5.6810.306.108.6716.22
Kathbath (noisy)6.9512.005.1010.1117.06
Common Voice7.6211.4010.3613.6123.55
Indic-TTS3.637.605.168.7510.72
MUCS3.2212.004.698.1516.20
Gramvaani15.6226.8020.8024.0931.44