Dictation glossary

Live partial transcription

Live partial transcription means showing in-progress text while audio is still being captured, instead of waiting until you stop talking to reveal a finished transcript. Streaming engines emit tokens as they go; batch engines wait for an endpoint. Partials can change: an early guess of 'wreck a nice beach' may settle into 'recognize speech' once more context arrives. The display is feedback, not always the final insert.

In more detail

What is Live partials?

Streaming versus batch is the core split. A streaming recognizer is built to update a hypothesis on every chunk of audio, which is how captions stay near live. A batch recognizer, including many Whisper-style pipelines, wants a complete-enough clip before it commits. Products can fake streaming by re-running a batch model on a growing buffer, which gives you partials without a native streaming decoder. Either way, the user-facing idea is the same: you see words before the session ends.

Partials are unstable by design. That flicker is useful — you catch a wrong name mid-sentence — and distracting if the HUD jumps too often. Latency of the first partial is a different number from latency of the finalized insert. Voice activity detection still matters: if the system does not know you started speaking, there is nothing to show. For dictation, live partials are a confidence display more than a legal transcript; the injected text should be the cleaned, final pass.

For writers who speak

Why it matters for dictation

Without partials, dictation is a leap of faith: you talk into a void and hope. With them, you steer. You stop early if the engine heard the wrong product name. You keep going if the sentence looks right. That feedback loop is what makes speaking feel like typing with a preview, rather than leaving a voicemail for your own editor.

In this product

How WhisperJot handles it

WhisperJot shows live partial text in a floating HUD while you dictate, including a live partial line during meetings. The inserted result is the finalized, cleaned transcript, not the flickering preview. Local engines are the default; Jot Cloud is opt-in and transcribes once at the end of a dictation, so there is no live partial text while it is selected.

Questions

Straight answers.

What is live partial transcription?

Live partial transcription is the in-progress text you see while you are still speaking. It updates as more audio arrives and may rewrite earlier words when the model grows more confident. The final transcript can differ from the last partial, especially after punctuation and filler removal. The feature exists so you have feedback instead of waiting in silence until the utterance ends.

How is streaming speech-to-text different from batch?

Streaming engines are designed to emit tokens continuously. Batch engines wait for a complete clip, file, or endpoint before producing a stable transcript. Some apps approximate streaming by repeatedly transcribing a growing recording with a batch model. Streaming usually shows faster first text; batch can be more accurate on a full sentence. Dictation products mix the two depending on the engine you pick.

Why do live partials change after I stop talking?

Because a partial is a hypothesis with incomplete audio. End-of-speech gives the model the rest of the phrase, and cleanup then deletes fillers and adds punctuation. Seeing a rewrite is normal, not a glitch. If the final insert is worse than the partial you liked, that is a cleanup or decoding issue, not proof that partials should be pasted raw.

One hotkey, any focused app.

Private local transcription by default, with an optional opt-in cloud engine.