Captions that don't need the venue's wifi.

No cloud. No per-minute billing.

Speech recognition runs on-device using Apple's speech engine. There is no cloud transcription service in the pipeline and no captioning traffic on your network — the only calls the app ever makes are a one-time OS model download and a license check.

After that, the venue's internet is irrelevant to your show.

Tracks speech instead of trailing it.

The engine streams partial words as they're spoken — real word-by-word delivery, not caption blocks landing a full clause behind the speaker.

Output is a normal macOS window, so it drops straight into the video path you already run: window capture in your switcher, a second output to a confidence monitor or IMAG screen, keyed into a stream layout.

Your client's look, not a template.

• Font, size, weight, spacing, alignment — auto-scaling with the window
• Colors, opacity, box width, anchored positioning for lower-thirds
• Per-character fade-in, configurable visible lines
• Named, saveable presets recalled instantly next show

All of it live in a preview pane while you adjust.

The failure modes, handled up front.

• Display never sleeps mid-session — quiet rooms included
• Mic disconnects detected immediately, remembered by hardware identity
• Stranded caption windows find their way back to a live display
• Tunable pause behavior — fast panels and deliberate keynotes both work
• One-click, content-free diagnostics safe to hand to anyone