Greedy decoding
Greedy decoding is the fastest way for a speech model to produce text: at every step it emits the single most probable next token, never revisiting earlier choices.
Greedy decoding is standard when speed matters — live captions, quick drafts — and its errors are usually local. Beam search reconsiders alternatives and wins on ambiguous audio at a compute cost.
Related terms
Frequently asked
What is Greedy decoding?
Greedy decoding is the fastest way for a speech model to produce text: at every step it emits the single most probable next token, never revisiting earlier choices.
Why does Greedy decoding matter?
Greedy decoding is standard when speed matters — live captions, quick drafts — and its errors are usually local. Beam search reconsiders alternatives and wins on ambiguous audio at a compute cost.
What terms are related to Greedy decoding?
Closely related concepts: Beam search, Streaming speech recognition — each has its own entry in this glossary.
Transcribe audio or video in 99+ languages — speaker labels, word timestamps, captions.
Transcribe a file →