Glossary · Speech
Is there a difference between ASR and STT?
ASR vs STT
ASR (automatic speech recognition) and STT (speech to text) are two names for the same job: turning spoken audio into written words. Researchers tend to say ASR, and API products tend to say STT.
On a call
A caller says "मुझे refund चाहिए, order last week आया था". The ASR model, sold to you as an STT API, writes that sentence down so the agent's language model can read it and reply.
AI agents from ₹1 a minute.¹
Explained
Here is how it works, in plain words.
If you’re comparing vendors, don’t let the two labels confuse you. A page selling “speech-to-text” and a paper about “ASR” are talking about the same thing.
The longer name came first. Automatic speech recognition has been a research field for decades, and textbooks still use it. “Speech to text” caught on with cloud APIs, because it says plainly what you send and what you get back. Jurafsky and Martin’s textbook uses both in one line: speech recognition, or speech-to-text, maps sound waves to written words.
What matters more is which kind of recognition you need.
Streaming recognition runs during a live call. It hears audio in small pieces and returns partial text as the caller speaks, then a final version when they stop. A voice agent depends on this, because the reply can’t start until the words exist.
Batch recognition, often sold as transcription, takes a finished recording. It can afford to listen to the whole thing, use more context, and add extras like who spoke when.
The two are scored the same way, usually with word error rate. They’re tuned differently. A streaming model trades a little accuracy for speed. A batch model trades speed for accuracy and polish.
For Indian calls, ask three questions of any recogniser. Does it handle Hindi and English in the same sentence? Was it tested on 8 kHz phone audio, or on clean studio speech? And how does it write English words inside Hindi: in Roman letters, the way people read them, or forced into Devanagari?
In Frequency
This is what Frequency does about it.
Fourier STT is our own speech recognition, built for Indian phone calls in Hindi, English, Gujarati, Tamil, Telugu and Hinglish. Our agents listen with it, and recorded calls can go through our transcription API, which returns speakers and timestamps.
Questions
People often ask us these questions.
Is transcription the same as ASR?
Mostly. Transcription often means the finished product for a recording, with speakers, timestamps and punctuation. ASR is the recognition step underneath it.
Still deciding? Ask us.
Fifteen minutes with the team that builds it. Bring your script, or your hardest call.
Book 15 minutes