Voice Session
Command Core
Reference arc-reactor
Primary interface
Voice and typed prompts use the protected J.A.R.V.I.S. backend, approval queue, scoped “Jarvis approved” flow, and spoken/browser speech response path so this screen can replace the external chat for day-to-day command.
Real-time voice
Command Core
Reference arc-reactor
Press to activate
Device capture scaffold
Requests browser microphone access only after Stufio taps the button. This local scaffold checks device support, captures an active stream, previews audio level, attempts browser-native STT, and keeps paid STT/TTS providers behind server-side token gates.
Secure context
missing
MediaDevices microphone
unknown
MediaRecorder scaffold
unknown
Browser STT
unknown
Native TTS playback
unknown
Active input
No microphone stream opened yet
Local level meter: 0%
STT transcript preview
idleTranscript preview will appear here when browser STT is available.
Safe local scaffold: nothing is uploaded, synthesized, stored, or sent to a provider.
Provider Adapter Readiness
This panel shows which provider adapters are selected, which environment variable names they expect, and which safety gates are closed. Token values stay server-side and are never rendered in the browser.
STT adapter
Server-side placeholder only. The adapter reports environment variable names and gate state, never token values, and it does not make live STT/TTS provider calls.
TTS adapter
Server-side placeholder only. The adapter reports environment variable names and gate state, never token values, and it does not make live STT/TTS provider calls.
Token exposure
blocked
Env reporting
names only
Live provider calls
not implemented
Fish Audio · Live TTS
Voice Configuration
Each agent in the command system has a dedicated voice identity slot. Configure ElevenLabs voice IDs here; no synthesis calls are made until an API key is present and a test is explicitly approved.
ElevenLabs provider
API key not configured
ElevenLabs API key is not yet configured. Add ELEVENLABS_API_KEY to the server environment to enable per-agent voice synthesis. No paid calls will be made until a key is present and a test is explicitly run.
Safety — paid API calls
ElevenLabs charges per character synthesized. This panel configures voice identity only. No synthesis request is made by loading this page, saving a voice ID, or viewing test status. A test call will be displayed as a destructive action requiring explicit confirmation before any API call is sent.
Provider-ready
ElevenLabs is the candidate for high-quality custom TTS. The app maps each agent to a voice ID slot. No voice provider API key is stored or requested here; credentials stay approval-gated until the provider and setup path are authorized.
Calm, authoritative commander
Central operating partner — primary voice for command layer, strategy, reporting
Precise, analytical
Engineering and architecture voice — CTO lane
Energized production director
Events, logistics, run-of-show execution voice
Creative and expansive
Creative strategy and innovation mapping voice
Unified comms
The app pulls the important feed and information into a clean command bowl, connected to tasks, approvals, rooms, revenue, and voice sessions — not a recreation of Slack.
Live Slack API ingestion is a separate approved integration decision. Current safe path: bridge notes and local command history first.
Final responses must not use a generic AI voice. Each agent needs a configured voice identity, STT/TTS handoff, fallback policy, and visible speaking state so the animated core reacts when the correct agent talks.