Glossary
Streaming ASR (also called online ASR) processes audio as it arrives, emitting partial transcripts continuously instead of waiting for the audio to finish. It's the technology behind live captions on Zoom, voice typing on phones, voice assistants, and live event captioning. Different architectural challenges than batch ASR — and a different accuracy/latency tradeoff.
Streaming vs batch architecture
Batch ASR (what we do) sees the full audio at once, can use bidirectional attention, and runs beam search with full lookahead. Streaming ASR sees audio in small chunks (50-200 ms) and must commit to text within ~1 second, before the rest of the sentence arrives. Architectures: RNN-Transducer (RNN-T), CTC-based models, transformer transducers, monotonic-attention transformers.
Why streaming is harder
Two main challenges. (1) Latency-accuracy tradeoff: shorter chunks = lower latency but less context = lower accuracy. Tuning this is a UX decision. (2) Partial result instability: "I scream" might appear and then get corrected to "ice cream" 200 ms later. UI/UX has to handle the rewriting gracefully — typically by italicizing or graying out unstable text until it's committed.
When you'd use streaming
Live captions (Zoom, Google Meet, Teams), voice assistants (Siri, Alexa, Google Assistant), live broadcast captioning (sports, news), dictation software, accessibility tools (real-time captions for deaf and hard-of-hearing in classrooms). Anywhere the user wants results during the speech rather than after. We don't currently offer streaming — our focus is post-recording batch transcription, where accuracy matters more than latency.
In practice
When you join a Zoom call with live captions enabled, Zoom is running a streaming ASR system on each speaker's audio. Words appear with ~1-2 second delay, sometimes flicker as the model corrects earlier guesses, and don't include speaker labels because diarization is hard in real time. After the meeting, re-process the recording in batch mode (uploading the .m4a to Transcript.you) for a more accurate transcript with speaker labels.
Related terms
Further reading
Transcribe a recorded meeting
Get started — freeLast updated: September 21, 2026
Frequently Asked Questions
What does Streaming ASR mean in transcription?
Speech recognition that emits partial results as audio arrives, instead of waiting for the recording to complete. Required for live captions and conversational AI.
Why does Streaming ASR matter when choosing a transcription workflow?
Streaming ASR affects how you understand transcript quality, timing, compatibility, or the technology behind speech-to-text results.
Where can I apply Streaming ASR on Transcript.you?
Use the related workflow at /zoom-to-text when you want to see how this glossary concept connects to an actual transcription task.
Browse the full transcription glossary or read the complete guide to AI transcription.