Whisper STT snippet

stt.py

#%pip install openai

from openai import OpenAI
import os
import subprocess
from pydub import AudioSegment
import math


client = OpenAI(api_key="sk-proj-XXX")

def transcribe_audio(file_path):
    audio_file = open(file_path, "rb")
    transcription = client.audio.transcriptions.create(
        model="whisper-1", 
        file=audio_file
    )
    return transcription

def convert_mp3_to_ogg(mp3_path, output_dir):
    """Convert an MP3 file to OGG format"""
    # Ensure output directory exists
    os.makedirs(output_dir, exist_ok=True)
    
    # Get the base filename without extension
    base_name = os.path.basename(mp3_path).rsplit('.', 1)[0]
    ogg_path = os.path.join(output_dir, f"{base_name}.ogg")
    
    # Convert MP3 to OGG using pydub
    audio = AudioSegment.from_mp3(mp3_path)
    audio.export(ogg_path, format="ogg")
    
    print(f"Converted {mp3_path} to {ogg_path}")
    return ogg_path

def split_audio_into_chunks(ogg_path, chunk_length_minutes, output_dir):
    """Split an OGG file into chunks of specified length"""
    # Ensure output directory exists
    os.makedirs(output_dir, exist_ok=True)
    
    # Load the audio file
    audio = AudioSegment.from_ogg(ogg_path)
    
    # Calculate chunk length in milliseconds
    chunk_length_ms = chunk_length_minutes * 60 * 1000
    
    # Calculate number of chunks
    total_chunks = math.ceil(len(audio) / chunk_length_ms)
    
    # Get the base filename without extension
    base_name = os.path.basename(ogg_path).rsplit('.', 1)[0]
    
    chunks = []
    # Split the audio and export chunks
    for i in range(total_chunks):
        start_ms = i * chunk_length_ms
        end_ms = min((i + 1) * chunk_length_ms, len(audio))
        
        chunk = audio[start_ms:end_ms]
        chunk_path = os.path.join(output_dir, f"{base_name}_chunk{i+1}.ogg")
        chunk.export(chunk_path, format="ogg")
        chunks.append(chunk_path)
        print(f"Created chunk {i+1}/{total_chunks}: {chunk_path}")
    
    return chunks

  

# Process the MP3 file
mp3_path = os.path.expanduser("~/Downloads/output_audio.mp3")  # Expand the tilde in the path
ogg_directory = "ogg/xx"
ogg_path = convert_mp3_to_ogg(mp3_path, "ogg")
chunks = split_audio_into_chunks(ogg_path, 15, ogg_directory)

print(f"Audio file split into {len(chunks)} chunks")

transcriptions = {}

for filename in os.listdir(ogg_directory):
    if filename.endswith(".ogg"):
        print(f"Processing {filename}")
        file_path = os.path.join(ogg_directory, filename)
        transcription = transcribe_audio(file_path)
        transcriptions[filename] = transcription