ONE✦FRIENDV2.0 · Private Validation

A voice that listens.

Animated neural companion · no face · user-controlled memory

Ready

Tap Voice to speak, or type a message below.

Voice recognition is turn-by-turn. No continuous background recording.

Voice level

Real-time voice (experimental)

WebRTC two-way audio. The microphone and AI audio are processed by the AI provider during a connected call. The ordinary chat mode below still works separately. Requires HTTPS, a deployed Worker, and a private test token. End Call fully closes the connection.

Not connected

Call time: 00:00 · private-beta limit: 05:00

Session: idle · Microphone: off · Connection: closed

I consent to sending my microphone audio to the AI service during live calls.

On iPhone, use Safari over HTTPS, keep the page foregrounded, and start the call with a tap. Network and device support vary. No automatic reconnect. Retry requires a new tap. Backgrounding ends the call.

Conversation

Demo mode uses scripted replies until the backend is configured. Interrupt stops voice playback; it does not support true simultaneous full-duplex voice. Voice recognition depends on Safari/browser support and may use the browser provider's speech services.

Private test settings

Companion namePersonalityLanguageSpeech voiceMicrophone sensitivitySpeaking speedModeBackend Worker URLPrivate test token (never stored) Automatically start the next listening turn after AI speaks (experimental; browser support varies) Save chat history on this device (opt-in) Include my approved memory notes in AI requests (opt-in)Approved memory notes, one per line

Available speech voices depend on the device and browser. Voice choices are stored only on this device.

V2.0 limits each experimental live call to five minutes on this device; this is a usability limit, not a billing cap. PRIVATE TEST ONLY — not independently audited. Do not enter confidential information. Saved chat and memory notes are stored in your browser without encryption. API chat messages and approved notes are transmitted to the configured backend. This is not yet approved for public ONECIRCLE users.

V2.0 release readiness

These checks run locally. They do not start a microphone or AI session, spend API credits, or certify iPhone compatibility.

Not checked

Manual device tests — mark these only after testing on the target iPhone and HTTPS deployment.

Microphone permission and rejection tested Live AI speech understood and audible Interrupt, mute and end-call tested Network loss, End Call while connecting, and reconnect tested API usage and billing guardrails reviewed Privacy and microphone release verified

NOT VERIFIED — live device testing required.