How to beginner · 3 min read

Whisper batch transcription

Q: Whisper batch transcription

Use the OpenAI Whisper API via the openai Python SDK to transcribe multiple audio files in batch by iterating over files and calling client.audio.transcriptions.create for each. Automate batch processing with a simple loop and handle results programmatically.

Quick answer

Use the OpenAI Whisper API via the openai Python SDK to transcribe multiple audio files in batch by iterating over files and calling client.audio.transcriptions.create for each. Automate batch processing with a simple loop and handle results programmatically.

PREREQUISITES

Python 3.8+
OpenAI API key (free tier works)
pip install openai>=1.0

Setup

Install the official openai Python package and set your OpenAI API key as an environment variable.

Install package: pip install openai
Set API key in your shell: export OPENAI_API_KEY='your_api_key' (Linux/macOS) or setx OPENAI_API_KEY "your_api_key" (Windows)

bash

pip install openai

Step by step

This example demonstrates batch transcription of multiple audio files using the OpenAI Whisper API. It loads each audio file, sends it to the whisper-1 model, and prints the transcribed text.

python

import os
from openai import OpenAI

client = OpenAI(api_key=os.environ["OPENAI_API_KEY"])

def batch_transcribe(file_paths):
    transcripts = {}
    for path in file_paths:
        with open(path, "rb") as audio_file:
            response = client.audio.transcriptions.create(
                model="whisper-1",
                file=audio_file
            )
            transcripts[path] = response.text
    return transcripts

if __name__ == "__main__":
    audio_files = ["audio1.mp3", "audio2.wav", "audio3.m4a"]
    results = batch_transcribe(audio_files)
    for file, text in results.items():
        print(f"Transcription for {file}:\n{text}\n")

output

Transcription for audio1.mp3:
Hello, this is the first audio file.

Transcription for audio2.wav:
This is the second audio file transcription.

Transcription for audio3.m4a:
Final audio file transcription text here.

Common variations

You can adapt batch transcription for asynchronous processing using asyncio to speed up multiple requests. Also, you can specify additional parameters like language or prompt to improve accuracy. Using different audio formats (mp3, wav, m4a) is supported.

python

import os
import asyncio
from openai import OpenAI

client = OpenAI(api_key=os.environ["OPENAI_API_KEY"])

async def transcribe_async(path):
    with open(path, "rb") as audio_file:
        response = await client.audio.transcriptions.acreate(
            model="whisper-1",
            file=audio_file
        )
        return path, response.text

async def batch_transcribe_async(file_paths):
    tasks = [transcribe_async(path) for path in file_paths]
    results = await asyncio.gather(*tasks)
    return dict(results)

if __name__ == "__main__":
    audio_files = ["audio1.mp3", "audio2.wav", "audio3.m4a"]
    results = asyncio.run(batch_transcribe_async(audio_files))
    for file, text in results.items():
        print(f"Async transcription for {file}:\n{text}\n")

output

Async transcription for audio1.mp3:
Hello, this is the first audio file.

Async transcription for audio2.wav:
This is the second audio file transcription.

Async transcription for audio3.m4a:
Final audio file transcription text here.

Troubleshooting

If you see FileNotFoundError: Verify the audio file paths are correct and accessible.
If you get InvalidRequestError: Check that the audio file format is supported and the file is not corrupted.
If transcription is inaccurate: Specify the language parameter if known, or provide a prompt to guide the model.
API rate limits: For large batches, add delays or use asynchronous calls to avoid hitting rate limits.

✅

Key Takeaways

Use the OpenAI Whisper API with client.audio.transcriptions.create to transcribe audio files one by one in batch.
Automate batch transcription by looping over audio files and collecting results in a dictionary.
For faster batch processing, use asynchronous calls with acreate and asyncio.
Specify optional parameters like language to improve transcription accuracy.
Handle common errors by verifying file paths, formats, and managing API rate limits.

Verified 2026-04 · whisper-1

Verify ↗