Whisper

Whisper (OpenAI) is the industry standard for Speech-to-Text. It supports 99 languages and translation. V3 (large-v3) is the current state of the art.

When to Use

• Transcription: Creating subtitles for videos.
• Translation: Translating audio to English text.
• Local Privacy: Runs 100% locally (sensitive meetings).

Core Concepts

Models

tiny, base, small, medium, large, legacy, large-v3, large-v3-turbo.

Distil-Whisper

Smaller, faster versions of Whisper (6x speedup, 1% accuracy loss).

Best Practices (2025)

Do:

• Use insanely-fast-whisper: A wrapper that uses Flash Attention to transcribing 2 hours of audio in 2 minutes.
• Use API for streaming: OpenAI API supports streaming audio transcription.

Don't:

• Don't use large for realtime: It's too slow. Use turbo or distil models.

References

• Whisper GitHub