v0.5.0 — speech-to-speech

Hands-free, interruptible voice conversation with 小光 — talk to her and she talks back, no clicking the mic.

Highlights

• 🍎 Apple on-device TTS — the new default. No API key, no network, fully private, works out of the box. Includes an in-app guide to downloading higher-quality system voices.
• 🎙️ Hands-free mode — the mic re-arms after each reply; just talk.
• 🗣️ Full-duplex voice barge-in (opt-in) — talk over her and she stops to listen, using on-device acoustic echo cancellation. It engages only while she's speaking, so other apps' audio isn't muted the rest of the time.

Also: a reader-friendly, bilingual (English + 繁體中文) README, with deep-dives moved into docs/.

📄 Voice setup · Full changelog

v0.4.0

👀 v0.4.0 — Screen vision

• Screen vision (opt-in, off by default) — let 小光 see what you're working on. She captures your focused window (or the whole screen) and sends it to your model, so you can ask "what's this error?" and she actually sees it. While it's on, she also glances at your screen periodically (configurable interval) and only speaks up when there's something worth mentioning. Enable under Settings → Screen Vision — a consent prompt + macOS Screen Recording permission are required, and it needs a vision-capable model.
• Pet mode keeps your focus — the desktop-pet overlay is now non-activating, so dragging her around no longer steals keyboard focus from your editor while she watches.

Full changelog: v0.3.0...v0.4.0

v0.3.0

🎙️ v0.3.0 — OpenAI TTS + pluggable speech-to-text

• OpenAI text-to-speech — route speech through OpenAI's /v1/audio/speech under Settings → Voice → TTS Provider → OpenAI, with promptable voice instructions, adjustable speed, and a Test Voice preview (works for every provider). Contributed by @canyugs (#3).
• Pluggable speech-to-text — pick your transcription provider under Settings → Speech Input: Apple on-device (default) or cloud Whisper via Groq, OpenAI, or any OpenAI-compatible endpoint. Contributed by @canyugs (#4).
• Docs + zh-Hant localization for the new settings, plus a note clarifying cloud vs on-device voice privacy.

Full changelog: v0.2.1...v0.3.0

v0.2.1

🐾 v0.2.1 — Desktop Pet speech bubble fix

• The pet-mode speech bubble moved beside her face. It used to sit at the bottom of the window, so when she was parked with her legs below the screen edge the text scrolled off-screen and you never saw it.
• The bubble now auto-flips to whichever side has more room, so it never clips the screen edge when she's near the left or right border.

Full changelog: v0.2.0...v0.2.1

v0.2.0

✨ What's new in v0.2.0

• 🐾 Desktop Pet mode — detach 小光 into a borderless, transparent, always-on-top overlay that lives on your desktop. Drag to move, scroll/pinch to resize, double-click to return to the window (🐾 toolbar button · Character menu · ⌘⇧D).
• 🎙️ Pluggable text-to-speech — choose your voice provider in Settings → Voice: cloud MiniMax, plus an experimental local BlueMagpie-TTS option (pending end-to-end verification). Contributed by @hlb.
• 🧍 Configurable character model — switch VRM models from Settings → Character instead of editing source. Contributed by @hlb.
• 🌐 Clearer language setting — the Language picker now notes that the interface language applies after an app restart (the character switches right away).

🙏 Thanks to @hlb for the BlueMagpie TTS provider and the configurable VRM model.

Full changelog: v0.1.0...v0.2.0