Command Explorer › gen: — what the agent makes › voice

gen:voice

A voice recording, made and delivered as an audio file.

Level 2Kind axisClass: actShortest: voRequires: generationPersistence: content-writeConfirmation in the reply: required

Directive Shorthand 3.5 · Command explorer · Assembled from the canonical spec, registry, usage guide, and implementation guide — this page adds no meanings.

What it does

A recording as the deliverable: narration, a voiceover, a spoken piece — synthesized speech produced and delivered as an audio artifact you can keep, replay, and send. The reply is not the recording; the recording is.

The registry's own words: A voice recording as a deliverable: narration, a voiceover, a spoken piece — synthesized speech produced and delivered as an audio artifact, distinct from a reply spoken aloud.

Fire it — usage examples

“Record the welcome message for new clients — warm, thirty seconds. gen:voice”

→ a voice recording is produced and delivered as an audio file, ready to send on.

“Take that announcement text and turn it into audio I can post. gen:vo”

→ the announcement is delivered as a recording. Contrast re:voice: that makes the reply itself spoken — say-it; this is make-it, a file.

Said aloud: “Lingo, generate voice — the announcement, as a recording.”

→ the recording is produced and delivered; the reply confirms the artifact — it does not become the audio.

Spoken. Inside a marked span opened by the wake word “Lingo”, the spoken form is “generate voice”. Dictated directives fire deliberately, never by accident.

What the agent actually does

The agent takes the script or the designated text from the message, synthesizes it in a fitting voice, and delivers the audio as an artifact — a file or playable attachment, not as the reply channel. Where a choice of voice or length is left open, it picks a reasonable one and says which; a host that cannot synthesize says so plainly rather than describing a recording it did not make.

The baseline procedure, its edge cases, and this directive's full guide live in the implementation guide: implementation/IMPLEMENTATION.md#cmd-gen-voice ↗.

When the host can't

No generation capability for voice on this host: disclose it plainly, never fake a recording, and never silently substitute a spoken reply or text for the artifact. A description of the recording is not the recording.

Conformance — “done when”

An audio artifact exists and is delivered (or the missing capability is disclosed) — never a description of a recording standing in for one.

Persistence class: content-write. Execution report in the reply: required. A user can test any of this by hand — fire the directive, then check the result against the store, the schedule, or the reply itself.