Automatic Speech Recognition (ASR)

Automatic Speech Recognition (ASR) is the task of transcribing spoken language into written text. It is the technology behind dictation, live captioning, voice assistants, and meeting transcripts. Early ASR systems stitched together separate acoustic, pronunciation, and language models, but modern ASR is dominated by end-to-end neural networks that map audio straight to text.

Whisper and the open-weight shift

The best-known modern ASR system is Whisper, released open-weight by OpenAI in 2022. Whisper uses a transformer encoder-decoder: the encoder ingests the audio signal and the decoder generates the transcript. Trained on more than five million hours of multilingual, multitask audio, it is unusually robust to accents, background noise, and technical jargon, and it handles transcription and translation across dozens of languages. It ships in several sizes (tiny through large) to fit different hardware budgets.

Why ASR matters for sovereignty

Because Whisper's weights are open, the entire transcription pipeline can run on your own machine, no cloud, no per-minute fees, and no audio leaving the device. For privacy-conscious users, self-hosted ASR means voice notes, interviews, and calls can be transcribed without trusting a third-party service. It is a clear example of how open weights turn a hosted convenience into a sovereign capability.

ASR is the input side of voice interfaces; its mirror image is covered in our text-to-speech entry, and both are built on the transformer architecture.