whisper-transcriber 0.1.0

A library for transcribing audio files using Whisper models

Whisper Transcriber

A Python library for transcribing audio files using Whisper models with intelligent silence detection and segmentation.

Installation

pip install whisper-transcriber

Requirements

• Python 3.7 or higher
• ffmpeg and ffprobe installed on your system

Features

• Intelligent silence detection for natural segmentation
• Adaptive audio analysis for optimal threshold detection
• High-quality transcription using Whisper models
• Support for various audio formats
• SRT subtitle output

Usage

Command Line

# Basic usage
whisper-transcribe audio_file.mp3

# Advanced usage
whisper-transcribe audio_file.mp3 --model loko99/whisper_small_kannada --output transcript.srt --min-segment 3 --max-segment 12

Python Library

from whisper_transcriber import WhisperTranscriber

# Initialize the transcriber
transcriber = WhisperTranscriber(model_name="openai/whisper-small", hf_token="YOUR_HF_TOKEN")

# Transcribe an audio file
results = transcriber.transcribe(
    "audio_file.mp3",
    output="transcript.srt",
    min_segment=5,
    max_segment=15,
    normalize_text=True
)

# Access the transcription results
for i, segment in enumerate(results):
    print(f"Segment {i+1}: {segment['transcript']}")

License

MIT