Whisper: Transcribe Audio to Text

whisper.py

# Sample script to use OpenAI Whisper API
# This script demonstrates how to convert input audio files to text, fur further processing.
# The code can be still improved and optimized in many ways. Feel free to modify and use it
# for your own needs.
#
import openai
from openai import OpenAI

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

import argparse
from rich import print
from rich.console import Console
from rich.traceback import install

# Set your API key

console = Console()
install()

def transcribe_audio(file_path, language):
    """Transcribe the audio using the Whisper API."""
    console.print(f"[bold blue]Starting transcription for file:[/bold blue] {file_path}")
    try:
        with open(file_path, "rb") as audio_file:
            console.print("[bold blue]Sending audio file to Whisper API...[/bold blue]")
            response = client.audio.transcribe(model="whisper-1",
            file=audio_file,
            response_format="json",
            language=language)
        console.print("[bold blue]Transcription received.[/bold blue]")
        if isinstance(response, dict) and "text" in response:
            return response.text
        else:
            console.print("[bold red]Unexpected response format received from Whisper API.[/bold red]")
            return None
    except openai.OpenAIError as e:
        if "Invalid language" in str(e) and "ISO-639-1 format" in str(e):
            console.print(f"[bold red]An API error occurred during transcription: Invalid language '{language}'. Language parameter must be specified in ISO-639-1 format.[/bold red]")
        else:
            console.print(f"[bold red]An API error occurred during transcription:[/bold red] {e}")
        return None
    except Exception as e:
        console.print(f"[bold red]An unexpected error occurred during transcription:[/bold red] {e}")
        return None

def summarize_text(text):
    """Generate a summary of the transcribed text using GPT-4."""
    console.print("[bold blue]Starting text summarization...[/bold blue]")
    try:
        summary = client.chat.completions.create(model="gpt-4",
        messages=[
            {"role": "system", "content": "You are a helpful assistant."},
            {"role": "user", "content": f"Summarize the following text:\n\n{text}"}
        ])
        console.print("[bold blue]Summary received.[/bold blue]")
        return summary.choices[0].message.content
    except openai.OpenAIError as e:
        console.print(f"[bold red]An API error occurred during summarization:[/bold red] {e}")
        return None
    except Exception as e:
        console.print(f"[bold red]An unexpected error occurred during summarization:[/bold red] { e }")
        return None

def generate_bullet_points(text):
    """Generate bullet points of key topics from the text."""
    console.print("[bold blue]Starting bullet points generation...[/bold blue]")
    try:
        bullet_points = client.chat.completions.create(model="gpt-4",
        messages=[
            {"role": "system", "content": "You are a helpful assistant."},
            {"role": "user", "content": f"Create bullet points from the following text:\n\n{text}"}
        ])
        console.print("[bold blue]Bullet points received.[/bold blue]")
        return bullet_points.choices[0].message.content
    except openai.OpenAIError as e:
        console.print(f"[bold red]An API error occurred while generating bullet points:[/bold red] {e}")
        return None
    except Exception as e:
        console.print(f"[bold red]An unexpected error occurred while generating bullet points:[/bold red] {e}")
        return None

def generate_mind_map_description(text):
    """Generate a mind map description based on the conversation."""
    console.print("[bold blue]Starting mind map description generation...[/bold blue]")
    try:
        mind_map = client.chat.completions.create(model="gpt-4",
        messages=[
            {"role": "system", "content": "You are a helpful assistant."},
            {"role": "user", "content": f"Describe how a mind map of the following text would look:\n\n{text}"}
        ])
        console.print("[bold blue]Mind map description received.[/bold blue]")
        return mind_map.choices[0].message.content
    except openai.OpenAIError as e:
        console.print(f"[bold red]An API error occurred while generating the mind map description:[/bold red] {e}")
        return None
    except Exception as e:
        console.print(f"[bold red]An unexpected error occurred while generating the mind map description:[/bold red] {e}")
        return None

def process_audio(file_path, options):
    """Process the audio file according to the specified options."""
    console.print(f"[bold blue]Processing audio file:[/bold blue] {file_path}")
    transcript = transcribe_audio(file_path, options.language)
    if not transcript:
        console.print("[bold yellow]No transcript available.[/bold yellow]")
        return

    if options.full_text:
        console.print("[bold green]Full Text:[/bold green]\n", transcript)

    if options.summary:
        console.print("[bold blue]Generating summary...[/bold blue]")
        summary = summarize_text(transcript)
        if summary:
            console.print("[bold green]Summary:[/bold green]\n", summary)

    if options.bullet_points:
        console.print("[bold blue]Generating bullet points...[/bold blue]")
        bullet_points = generate_bullet_points(transcript)
        if bullet_points:
            console.print("[bold green]Bullet Points:[/bold green]\n", bullet_points)

    if options.mind_map:
        console.print("[bold blue]Generating mind map description...[/bold blue]")
        mind_map = generate_mind_map_description(transcript)
        if mind_map:
            console.print("[bold green]Mind Map Description:[/bold green]\n", mind_map)

def main():
    parser = argparse.ArgumentParser(description="Transcribe audio and generate various outputs using OpenAI Whisper API.")
    parser.add_argument("audio_file", help="Path to the input audio file (mp3 format).")
    parser.add_argument("--summary", action="store_true", help="Generate a summary of the conversation.")
    parser.add_argument("--bullet_points", action="store_true", help="Generate bullet points from the most talked about topics.")
    parser.add_argument("--mind_map", action="store_true", help="Generate a mind-map graph of subjects discussed during the conversation.")
    parser.add_argument("--full_text", action="store_true", help="Get the full text of the conversation.")
    parser.add_argument("--language", default="en", help="Specify the language of the audio for transcription (default: English). Language must be in ISO-639-1 format.")

    args = parser.parse_args()

    process_audio(args.audio_file, args)

if __name__ == "__main__":
    main()