OpenAI Whisper

OpenAI Whisper is an open-source automatic speech recognition (ASR) model that transcribes audio to text with high accuracy across 100+ languages, including automatic language detection and translation to English. It has become the de facto standard for AI-powered transcription.

Key use cases include:

• Meeting and call transcription for business productivity tools
• Podcast and video subtitle generation
• Voice-to-text for mobile and desktop applications
• Multilingual audio transcription and translation
• Building voice-enabled AI assistants (speech-to-text pipeline stage)
• Media indexing and searchable audio archives

Whisper is used by developers building any application that needs to convert audio to text. It can be run locally (open-source) or via the OpenAI API. The open-source nature has spawned optimized variants like faster-whisper and whisper.cpp that run 4-8x faster.

Whisper handles challenging audio well — background noise, accents, multiple speakers, and technical terminology are all handled with strong accuracy.

1. Via OpenAI API (simplest):

   pip install openai
   
   from openai import OpenAI
   client = OpenAI()
   transcription = client.audio.transcriptions.create(
       model='whisper-1',
       file=open('audio.mp3', 'rb')
   )
   print(transcription.text)

2. Self-hosted (free, requires GPU):

   pip install openai-whisper
   whisper audio.mp3 --model large-v3

3. Optimized self-hosted (recommended):

   pip install faster-whisper
   from faster_whisper import WhisperModel
   model = WhisperModel('large-v3', compute_type='float16')
   segments, info = model.transcribe('audio.mp3')

Pricing: OpenAI API charges $0.006/minute of audio. Self-hosted is free but requires a CUDA GPU for reasonable speed (large-v3 needs ~10GB VRAM). CPU inference is possible but significantly slower.