Skip to main content

Shisa ASR

Shisa ASR is a speech recognition (speech-to-text) API tuned for Japanese, fluent in English, and built for production workloads. It converts spoken audio into accurate text with a single JSON request, can detect the spoken language automatically, and lets you boost domain-specific terms with custom vocabulary.

POST https://api.shisa.ai/asr/srt/audio_llm

Send base64-encoded audio (WAV, OGG, MP3, or FLAC) and receive the transcription, the detected language, and a confidence score. Only the audio field is required — language is auto-detected and tuning parameters use sensible defaults.

Why Shisa ASR​

  • Japanese-first design — tuned for Japanese, including regional speech, and fluent in English.
  • Real-time streaming — process audio streams for live transcription and instant results.
  • Voice activity detection — keep one transcript with VAD on or split batch audio into timestamped speech segments.
  • Enterprise security — see the Privacy Policy on the Shisa platform for how your data is handled.
  • Custom vocabulary — add industry-specific terms, brand names, and custom phrases via hotwords for better accuracy.

Use cases​

  • Call-center analytics — automatically transcribe customer service calls for quality assurance, compliance, agent training, and customer sentiment analysis.
  • Meeting notes — transform meetings, interviews, and discussions into searchable, actionable text documents.
  • Subtitles & captions — generate accurate subtitles for videos, live streams, and broadcasts in real-time or batch mode.

Next steps​

  • Quickstart — make your first transcription with curl, Python, or JavaScript.
  • API reference — request parameters, response fields, and error handling.
  • Audio & languages — supported formats and language-selection behavior.
  • Pricing — how ASR usage is billed.