1.0.4

Dictation now defaults to text appearing after each spoken phrase, with the recording saved as a note attachment. Existing saved choices remain unchanged; fresh installations and missing settings use the new defaults.

Voice selection separates the service, model, search and result sections. Configuration dialogs and settings wrap controls on narrow screens, voice buttons accommodate longer labels, and keyboard focus uses neutral colors. Floating reading and dictation bars wrap and remain separated when both are visible.

Validated with automated tests and real macOS Obsidian checks across reading, dictation, About, voice search, transcription service setup and first-use selection. Narrow layouts were simulated on desktop; mobile hardware remains unverified.

1.0.2

Mac system speech automatically discovers installed voices, including enhanced voices. Model changes now update applicable voices; searchable voice selection and audition preserve your reading position.

Adding a transcription service opens a configuration dialog. Cancel leaves your settings unchanged; confirmed duplicate configurations reuse the existing service. API key application links remain available in the dialog.

Qiaomu AI RSS can bundle the same speech engine independently. When both plugins are installed, starting or previewing speech pauses the other plugin; settings can be copied explicitly once, and local model data is shared. Neither plugin is required by the other.

Validation: automated provider/model, native-voice and preview restoration tests, plus real Obsidian checks for service setup and native Mac voices. Cloud connection tests require your own service account.

1.0.1

Read any note aloud and dictate straight into notes. Read the selected text, start at your cursor, or follow the visible reading view. Pause, resume, change speed and voices, and keep reading highlights in sync.

Cloud TTS and speech-to-text settings include clickable API-key application links and setup guides. Required fields stay visible, keys can be shown or hidden, and successful tests are cleared after settings change. ASR keys are isolated by provider; cancelled or outdated connection tests cannot report success.

Cloud recognition supports SiliconFlow, Doubao, GLM, StepFun, Xiaomi MiMo, Groq and OpenAI. Local SenseVoice, Kokoro and MeloTTS use an optional external runtime; the plugin downloads model data only. System speech, Mac speech and Edge TTS need no API key. Failed dictation retains the recording for retry or saving.

English and Simplified Chinese interfaces. Desktop-only local features are kept separate from mobile cloud/system speech. macOS Obsidian host checks and automated tests are recorded in the repository; Windows and Linux CI do not establish device playback, and mobile device testing is not yet complete.