Kodama JA Streaming ASR

A Japanese streaming automatic speech recognition (ASR) demo using ayousanz/kodama-ja-streaming-small, a full fine-tune of moonshine-ai/moonshine-streaming-small (MIT, English) on 35,000 hours of ReazonSpeech v2.

How to use

1. Upload or record a Japanese audio clip (16 kHz mono recommended).
2. Click Transcribe to get the Japanese text transcription.

Example audio

The example audio clips are from the Kokoro Speech Dataset, which is in the public domain.