Typhoon ASR

A user-friendly Python package for Thai speech recognition using the Typhoon ASR model.

Installation

pip install typhoon-asr

Quick Start

Python API

from typhoon_asr import transcribe

# Basic transcription
result = transcribe("audio.wav")
print(result['text'])

# With word timestamps (estimated)
result = transcribe("audio.wav", with_timestamps=True)
print(result['text'])
for ts in result['timestamps']:
    print(f"[{ts['start']:.2f}s - {ts['end']:.2f}s] {ts['word']}")

# Specify device
result = transcribe("audio.wav", device="cuda")  # or "cpu", "auto"

Command Line

# Basic usage
typhoon-asr audio.wav

# With timestamps
typhoon-asr audio.wav --with-timestamps

# Specify device
typhoon-asr audio.wav --device cuda

Supported Formats

• .wav, .mp3, .m4a, .flac, .ogg, .aac, .webm

API Reference

transcribe(input_file, model_name="scb10x/typhoon-asr-realtime", with_timestamps=False, device="auto")

Parameters:

• input_file (str): Path to audio file
• model_name (str): HuggingFace model identifier
• with_timestamps (bool): Generate estimated word timestamps
• device (str): Processing device ("auto", "cpu", "cuda")

Returns:

• dict with keys:
  • text: Transcribed text
  • timestamps: List of word timestamps (if enabled)
  • processing_time: Processing duration in seconds
  • audio_duration: Audio duration in seconds

Requirements

• Python ≥ 3.8
• CUDA (optional, for GPU acceleration)

License

Apache Software License 2.0