Dictations

What it is

Dictation in Omniscio works at the operating-system level: you hold a hotkey, speak, and the cleaned transcript is typed wherever your cursor is — in Omniscio or any other app. Every one of those dictations leaves two things behind: the words the engine heard, and a durable copy of the captured audio on your computer. Until this hub existed, those two lived in separate lists with nothing linking them, so a dictation whose transcription failed showed up as audio with no words in one list and (if it had any) words with no audio in the other.

The Dictations hub is that single list. One row per dictation, the words beside the recording they came from, newest first. It is also where you go to recover a dictation that was cut off.

Where to find it

Open it from any of:

• the floating dictation pill's History button,
• Open Dictations at the top of Settings → Voice Control → FlowVoice → History,
• the sidebar, under the Communication group,
• the command palette, by typing "Open Dictations".

On the whole, the hub is always present — there is no setting that hides it.

How it behaves

What each row shows

• When you dictated, how long it ran, and which app was in front.
• The words the engine produced — the cleaned text, or the raw transcript if cleanup was off.
• A status: Transcribed (the ordinary case), Recovered, May be incomplete, or Not transcribed yet when a recording never produced words.

What you can do on a row

• Play — hears the recording back. The audio is read from this computer only when you press Play; nothing is uploaded. Playback is desktop-only, because the audio file lives on this computer.
• Copy — puts the dictated words on your clipboard.
• Transcribe again — re-runs transcription on the saved audio (a fresh session of the same engine, then the built-in offline transcriber) and copies the result. This is how a recording that failed the first time gets its words back.
• Delete — permanently discards that recording's audio. Any text already in your history stays.

Needs attention

The list leads with the dictations that want a look — a recording that was never transcribed, or one flagged as possibly incomplete. Show all dictations reveals everything, ordinary ones included. Nothing is hidden because of its status; the filter only decides what leads.

A dictation a restart cut off

If Omniscio quits, crashes or restarts while you are mid-sentence, the recording survives: shutdown finishes writing it, and the next start transcribes it for you. You get one inbox card per recording, saying either that it was recovered (the words are now readable in the hub) or that its audio is saved and can be transcribed again. The card's Open dictations button opens this hub focused on that recording, highlighted, so you never have to hunt for it. The dictated words do not appear on the card itself, because an inbox alert can be forwarded to a teammate.

On a phone

The list, Copy and Transcribe again work in the mobile web app. Play does not — the audio lives on your computer, so there is nothing for the phone to read. Dictation itself is desktop-only.

For agents

The hub reads the flowvoice:dictations-list channel and plays audio through flowvoice:recordings-audio; both are backed by the same stores the settings card uses.

GET /flowvoice/dictations answers the same merged list, newest first. A row carries what you SAID, so the route is gated on the machine-wide CLI token (~/.amc/cli-token) rather than a scoped in-app-session token — the metadata-only GET /flowvoice/recordings still answers with plain read access.

Related

• FlowVoice (OS-wide dictation) — the hotkeys, the speech engines and what they cost, and where to configure it.
• FlowVoice — part 2 — what happens to your recording when a dictation fails.
• Voice History — in-app and per-session voice conversations, a different feature.