Mickai · SIOS voice subsystem
Mickai Vinis™
Mickai Vinis is the voice subsystem of the Mickai Sovereign Intelligence Operating System. Voice-biometric speaker verification, fresh-utterance authority on high-impact actions, on-device F5-TTS voice cloning, and a signed actuator that can drive the entire computer by voice.
View capabilities
Voice authority, sealed
The spoken word, made load-bearing.
Vinis verifies the speaker, demands a fresh utterance for any high-impact action, and signs every gate decision into the Open Audit Record. Capture, verification, and synthesis happen on the host. The audio never leaves the machine.
Deterministic invocation
A fresh voice, then the action runs.
On the record
Filed, not narrated.
Vinis sits on three of the 101 filed UK patent applications behind the Mickai SIOS, approximately 2,234 claims, owned by Mickai LTD, named inventor Micky Irons.
The Mickai SIOS
Mickai is a Sovereign Intelligence Operating System (SIOS). It runs entirely on your own hardware, on Windows, Linux, or macOS. No cloud, no telemetry. This page describes one subsystem of the Mickai SIOS. Request a key to install on your hardware.
A subsystem of the Mickai SIOS. Voice gating, voice-biometric authentication, speech synthesis, voice cloning via the F5-TTS sidecar, and a signed actuator that can drive the entire computer by voice. Actor identity cannot be supplied by a public reply.
Voice as a load-bearing authority signal. Speaker verification, fresh-utterance authority, on-device cloning, and a signed actuator that can drive the entire computer. Audio never leaves the machine.
What Vinis does
Eight primitives that turn voice into a load-bearing authority signal and a control surface for the entire workstation. Every gate decision and every actuator call is signed into the Open Audit Record substrate. Audio is captured, verified, and rendered entirely on the host.
01 / Speaker
Voice-biometric speaker verification
Continuous speaker verification on every utterance. Enrolled voiceprints are stored on-device and matched against captured audio without ever leaving the host. Fails closed: an unrecognised speaker cannot drive the actuator.
02 / Authority
Fresh-utterance authority on high-impact actions
Destructive or load-bearing tool invocations require a fresh, recently-spoken utterance bound to the actor. Replay attacks fail by construction. The fresh-utterance window is configurable per skill and per environment.
03 / Gate
Voice-gated deterministic tool invocation
Every tool call from the agent is gated by voice. The gate decision is signed into the Open Audit Record substrate so a regulator can replay the authority chain offline.
04 / STT
On-device speech-to-text
Whisper / Sherpa-ONNX run locally. Transcripts are signed and committed to the audit ledger before downstream tools see them. The transcript is the evidence.
05 / TTS
On-device speech synthesis and voice cloning
F5-TTS sidecar generates synthesised speech on the host. Cloned voices are watermarked under the AudioSeal dual-layer scheme so any output remains attributable.
06 / Watermark
AudioSeal dual-layer watermarking
Two independent watermarks are embedded in every synthesised output: a robust steganographic mark and a fragile cryptographic signature. Tampering is detectable; provenance is preserved.
07 / Actuator
Signed full-computer voice actuator
A voice command can drive the desktop, the shell, and the browser. Every actuator call is signed at commit so the action chain is reconstructable end to end.
08 / Sovereignty
Audio never leaves the machine
Capture, verification, transcription, synthesis, and cloning all run on the host. No third-party speech endpoints, no cloud-side audio retention, no foreign processing.
Patent anchors
Vinis sits on three of the 101 filed UK patent applications behind the Mickai SIOS. Patent 13 anchors voice-gated deterministic tool invocation; patent 06 covers extreme-environment speaker verification; patent 11 covers AudioSeal dual-layer watermarking on synthesised output.
- 13Voice-gated deterministic tool invocation
- 06Extreme-environment speaker verification
- 11AudioSeal dual-layer watermark on synthesised output
GB2607309.8 to GB2611702.8, GB2611885.1 onwards, and GB2612762.1 to GB2612793.6 · 101 filed UK patent applications · Approximately 2,234 claims
Wired with
- F5-TTS sidecar (on-device voice cloning)
- Whisper / Sherpa-ONNX (on-device STT)
- Resemblyzer-style speaker embeddings
- AudioSeal dual-layer watermarking
- ML-DSA-65 signed actuator commits (FIPS 204)
- Open Audit Record (OAR) emit pipeline
- TPM 2.0 / secure-enclave key custody
- WebAudio capture in the browser surface
- Native CoreAudio / WASAPI capture on desktop
- Voice-print enrolment ceremony with hardware attestation
- Per-skill clearance gating
- Fresh-utterance authority window per skill
Read the Vinis patents.
The Vinis voice authority is filed as part of 101 UK patent applications behind the Mickai SIOS. Read the claim text on the patent surface, or request an access key and enrol your voiceprint on a sovereign host yourself.
Engineered by Mickai LTD, United Kingdom · @mickyirons