SpeechRecognition

// Overview

SpeechRecognition is a long-running Python library that gives you a single interface for turning audio into text, whichever engine is behind it. You write the capture and recognition code once, then choose — or switch — between offline and hosted backends without restructuring your program.

The offline options are CMU Sphinx, the Vosk API, OpenAI Whisper running locally, Tensorflow and Snowboy hotword detection. The online options include Google Speech Recognition and Google Cloud Speech, Microsoft Azure Speech, IBM Speech to Text, Wit.ai, Houndify, the OpenAI transcription API, Groq's Whisper API and Cohere Transcribe. That spread is the point: prototype against a free hosted endpoint, then move to a local Whisper or Vosk model when privacy, cost or latency demands it.

It also handles the unglamorous part — microphone capture and audio-file loading — which is why it remains a common first dependency for voice projects. It is maintained by Anthony Zhang (Uberi), requires Python 3.10 or newer, and is released under the 3-clause BSD licence.

// What it does

• One recognition API across many backends, so swapping engines is a call change rather than a rewrite
• Offline engines: CMU Sphinx, Vosk, local OpenAI Whisper, Tensorflow, and Snowboy hotword detection
• Online engines: Google Speech Recognition, Google Cloud Speech, Azure Speech, IBM Speech to Text, Wit.ai, Houndify, OpenAI, Groq Whisper and Cohere Transcribe
• Built-in microphone capture as well as audio-file input
• Permissive 3-clause BSD licence, suitable for commercial use
• Ships runnable examples, including microphone_recognition.py and audio_transcribe.py

// Getting started

Install from PyPI and run the built-in module to confirm your microphone and defaults work before writing any code. Python 3.10+ is required.

pip install SpeechRecognition

python -m speech_recognition

Commands and code are distilled from the project's own documentation — always check the official repo for the latest.

// When to use it

• Reach for it when prototyping a voice feature and you do not yet know which recognition engine you will ship with
• Reach for it to move an existing prototype from a hosted API to local Whisper or Vosk without rewriting the audio layer
• Reach for it for batch transcription of recorded audio files in a Python pipeline
• Reach for it when you need microphone capture handled for you rather than dealing with PortAudio directly

// How SpeechRecognition compares

SpeechRecognition alongside other open-source audio, music & voice tools AI/TLDR tracks, ranked by GitHub stars.