Speech recognition
lastmod 2026-10-30created 2019-01-0918 edits
Speech recognition (automatic speech recognition (ASR), computer speech recognition, or speech-to-text (STT)) is a sub-field of computational linguistics concerned with methods and technologies that translate spoken language into text or other interpretable forms.
Speech recognition is the process by which a computer interprets spoken human language and converts its content into text data. It is also called STT (Speech-to-Text). It is drawing attention as a way to enter text in place of a keyboard. It is applied where device control and information retrieval by voice are needed, such as in robots and telematics. The representative algorithm is the HMM (Hidden Markov Model), which builds an acoustic model by statistically modeling speech uttered by various speakers and builds a language model through corpus collection.
It is also used for purposes such as personal authentication by comparing against previously recorded voice patterns; this is called speaker recognition.
Categories
- Speech recognition (Speech-to-Text; STT)
- Speech synthesis (Text-to-Speech; TTS)
- Voice Changer
- Automatic speech recognition (ASR)
- Spoken language identification (LID)
- Speech emotion recognition (SER)
- Audio event detection (AED)
- meeting_minutes_pipeline.py
Project
- Kaldi
- Omni SenseVoice - fast speech recognition with per-word timestamps
- Voice-Pro - Gradio WebUI supporting speech recognition, translation, and speech synthesis
- Shezem-rs - Rust-based fast audio fingerprinting system
- RealtimeVoiceChat - real-time AI voice chat with about 500 ms latency
- Caret - Caret 2.0: more accurate conversation transcripts, mobile app
- Voxtral transcribes - Mistral's speech-to-text model
- FreeFlow - open-source voice dictation app
- Moonshine - open-source fast, high-accuracy speech recognition (ASR) for edge devices
- Nightingale - open-source karaoke app that can use any song on your computer
- Cohere Transcribe - SOTA open-source speech recognition model
- Whisper
- WhisperX
- NVIDIA Nemotron 3.5 Transcribe - open speech recognition model supporting 40 languages, including Korean