Opus is an open, royalty-free audio codec designed for interactive speech and music over the internet. It is the default codec for WebRTC and much of modern real-time voice, valued for low latency and good quality across a wide range of bitrates. Because so much live audio travels as Opus, it is often the format that reaches a speech-to-text system.
History and standardization of Opus
Opus became an IETF standard in 2012, published as RFC 6716. It grew out of a merger of two projects: Skype's SILK coder and Xiph.Org's CELT coder, brought together through the standards process so the result would be a single open format rather than a proprietary one.
The royalty-free licensing was deliberate, aimed at a codec anyone could ship without patent fees, and that openness is a large part of why browsers and real-time frameworks adopted it so widely.
How Opus works
Opus is a hybrid codec. It combines SILK, a speech-oriented coder derived from linear prediction, with CELT, a transform coder that handles music and higher-fidelity audio, and it switches or blends between them based on the content.
Opus adapts its bitrate from around 6 kbps up to 510 kbps and supports sample rates from narrowband 8 kHz to fullband 48 kHz. Its low algorithmic delay, configurable down to a few milliseconds, suits two-way conversation, and it degrades gracefully by using packet-loss concealment to mask dropped audio on unreliable networks.
Why Opus matters for speech-to-text
Voice-agent and contact-center audio frequently arrives as Opus over WebRTC or SIP. At the low bitrates typical of calls, Opus preserves speech intelligibility better than older telephony codecs such as G.711, which makes it a sound default when you control the path.
Opus is still lossy, though: every encode discards some detail, so needless transcoding between formats stacks the loss and can quietly raise error rates.
When you decode Opus for recognition, match the sample rate the model expects and avoid extra hops, since the codec and the bitrate it runs at set a ceiling on the accuracy any downstream model can reach.
Frequently asked questions
What is the Opus codec? It is an open, royalty-free audio codec standardized as IETF RFC 6716, built for low-latency speech and music streaming. It powers WebRTC and most browser-based real-time voice.
Is Opus good for speech and voice calls? Yes. Its SILK component is tuned for speech, and at call bitrates it keeps voice clearer than G.711, while its low delay and packet-loss concealment suit live two-way audio.
Can speech-to-text transcribe Opus audio? Yes, once the audio is decoded to PCM at the rate the model expects. Most providers accept Opus directly; the practical caution is to avoid transcoding it through other lossy formats first.
How is Opus different from MP3 or G.711? Opus is newer and hybrid, covering both speech and music from very low to high bitrates with low latency. G.711 is a fixed narrowband telephony codec, and MP3 targets music rather than interactive voice.