title: Specifications description: Implementation-ready designs: what gets built, and how it is judged done. Part of the YazSes engineering and science documentation.
Specifications¶
Implementation-ready designs: what gets built, and how it is judged done.
This is the engineering tier
These pages are written for contributors, reviewers and researchers rather than for people using YazSes. For the user documentation, start at Home.
22 documents.
- Spec: First-class Mandarin language support
- Spec: Cocktail Filter — Target-Speaker Gate & Suppression
- Spec: Diarized Recording Import —
yazses transcribe <file> - Spec: Face-Gesture Switch — Calibrated Deliberate Expression as an Accessibility Input
- Spec: Grounded Targets — coarse pointing to exact semantic UI entities
- Spec: Hands-Free Bundle — Compose Voice, Gaze, Head Pointer and Switch Inputs Safely
- Spec: Head-Pointer Runtime — Pose to Pointer with Recenter, Clutch and Dwell
- Spec: Implicit Gaze Calibration — Click Association, Held-Out Promotion and Topology Safety
- Spec: Pointer Output — Cross-Platform Motion, Click and Scroll Boundary
- Spec: Shared Camera Perception — One Webcam, Derived Signals, Multiple Consumers
- Spec: Ghost Ahead — Endpoint / Turn Anticipation (latency hiding)
- Spec: Glance-Type — Coarse Gaze-Targeted Dictation
- Spec: Mid-Thought Undo — Spoken Buffer Revision
- Persian Benchmark and Validation Specification
- Persian (فارسی) Language Support Program
- Persian Text and RTL Specification
- Spec: Polyglot Switch — Configured-Pair Code-Switching ASR
- Spec: Prosody Ink — prosody-driven text formatting
- Spec: Punch-In — re-speak-to-correct span splicing
- Spec: Read-Back Loop — eyes-free dictation with offline TTS confirm/correct
- Spec: Say-Macro — User-Programmable Voice Launcher
- Spec: Voiceprint Mind — On-Device Personal Adaptation