"""
Whisper Service Module
Handles OpenAI Whisper model loading, caching, and audio transcription functionality.
"""
import whisper
import tempfile
from pathlib import Path
from config.settings import settings
from config.constants import (
DEFAULT_WHISPER_MODEL,
WHISPER_MODEL_NAMES,
TEMP_FILE_SETTINGS,
ERROR_MESSAGES,
SUCCESS_MESSAGES
)
class WhisperService:
"""Service class for handling Whisper model operations"""
def __init__(self, model_name=None):
"""
Initialize WhisperService
Args:
model_name (str): Whisper model name (tiny, base, small, medium, large)
If None, uses configuration setting
"""
self.model_name = model_name or settings.whisper_model
self.model = None
# Validate model name
if self.model_name not in WHISPER_MODEL_NAMES:
print(f"Warning: Invalid model name '{self.model_name}'. Using default '{DEFAULT_WHISPER_MODEL}'")
self.model_name = DEFAULT_WHISPER_MODEL
def load_model(self):
"""
Load the Whisper model once and cache it
Returns:
whisper.Whisper: Loaded Whisper model
"""
if self.model is None:
print(f"Loading Whisper model ({self.model_name})...")
self.model = whisper.load_model(self.model_name)
print("Model loaded successfully.")
return self.model
def transcribe_audio(self, audio_path):
"""
Transcribe an audio file using the loaded Whisper model
Args:
audio_path (str): Path to the audio file
Returns:
tuple: (transcription_text, temp_file_path) or (error_message, None)
"""
if not audio_path:
return "Please provide an audio file path.", None
try:
# Load model if not already loaded
whisper_model = self.load_model()
print(f"Starting transcription for '{audio_path}'...")
# Transcribe the audio
result = whisper_model.transcribe(audio_path, fp16=False)
transcription_text = result["text"].strip()
# Create a temporary text file for download
temp_file = tempfile.NamedTemporaryFile(
mode='w',
suffix='.txt',
delete=False,
encoding='utf-8'
)
temp_file.write(transcription_text)
temp_file.close()
return transcription_text, temp_file.name
except Exception as e:
error_msg = f"An error occurred during transcription: {str(e)}"
print(error_msg)
return error_msg, None
def transcribe_for_gradio(self, audio_file):
"""
Transcribe audio specifically for Gradio interface
Args:
audio_file: Gradio audio file object or path
Returns:
tuple: (transcription_text, temp_file_path) or (error_message, None)
"""
if audio_file is None:
return "Please upload an audio file.", None
# Get the file path from the uploaded file
audio_path = audio_file.name if hasattr(audio_file, 'name') else audio_file
return self.transcribe_audio(audio_path)
# Global instance for backward compatibility
_whisper_service = WhisperService()
def load_whisper_model():
"""
Legacy function for backward compatibility
Load the Whisper model once and cache it
"""
return _whisper_service.load_model()
def transcribe_audio_gradio(audio_file):
"""
Legacy function for backward compatibility
Transcribes an audio file using OpenAI's Whisper model for Gradio interface.
"""
return _whisper_service.transcribe_for_gradio(audio_file)
