- Overview
- Scorecard
- Updates19
Voice recording → AI polish → Insert into notes. One-step voice writing workflow.
Record your voice, let AI polish the text, and insert it directly at your cursor position or create a new note. Supports multiple speech-to-text providers and mixed Chinese/English recognition.
Features
- 🎙 Voice Recording — Floating draggable panel, non-blocking UI
- 🤖 Multiple STT Providers — Volcengine / Qwen (DashScope)
- ✨ AI Text Polishing — DeepSeek integration, 4 styles (formal/concise/casual/raw)
- 📝 Insert Mode — Text goes wherever your cursor is (any input field)
- 🎵 Note Mode — Auto-create date-organized notes with playable audio
- 📱 Mobile Support — Works on iOS and Android
- 🌐 Bilingual — Chinese and English speech recognition
功能
- 🎙 语音录制 — 浮动录音条不阻塞操作，可拖到任意位置
- 🤖 多 STT 服务商 — 支持火山引擎 / 千问（通义 DashScope）
- ✨ AI 润色 — 对接 DeepSeek，支持正式/简洁/口语化/仅纠错四种风格
- 📝 插入模式 — 光标在哪个输入框，文字就插哪，不限于 Obsidian 编辑器
- 🎵 笔记模式 — 自动创建带日期文件夹的笔记 + 可播放的录音音频
- 📱 移动端支持 — 安卓/iOS 均可使用
- 🌐 中英文识别 — 中英文混说也能正确转录
Usage
- Click the 💬 icon in the ribbon bar to open the floating recorder
- Click 🎤 to start recording, click ⏹ to stop
- Speech is automatically transcribed, polished, and inserted
Mode Switching
Click the 📝/🎵 button on the floating panel to switch:
| Mode | Icon | Behavior |
|---|---|---|
| Insert | 📝 | Text inserted at cursor position |
| Note | 🎵 | Creates a new markdown note with embedded audio |
Installation
From Community Plugins (pending review)
Search "AI Voice Polish" in Obsidian's community plugins browser.
Manual Installation
Download main.js, manifest.json, and styles.css from the latest release and copy them to your vault's .obsidian/plugins/ai-voice-polish/ folder.
Configuration
STT Provider
| Provider | Required Config |
|---|---|
| Volcengine | Access Key + Secret Key + App ID |
| Qwen (DashScope) | API Key |
Text Polishing
- DeepSeek API Key
- Polish style: formal / concise / casual / raw
- Custom prompt instructions
File Paths
- Notes folder: where
.mdfiles are stored - Audio folder: where
.webm/.mp4files are stored
Both support date-based subfolder organization.
Build
npm install
npm run build
License
MIT
HealthExcellent
ReviewSatisfactory
About
Record voice with a floating, movable recorder and transcribe mixed Chinese/English speech via multiple STT providers. Polish transcriptions with AI in formal, concise, casual or correction-only styles, then insert text at the cursor or create dated Markdown notes with embedded playable audio.Details
Current version
0.1.19
Last updated
2 months ago
Created
2 months ago
Updates
19 releases
Downloads
152
Compatible with
Obsidian 1.5.0+
Platforms
Desktop, Mobile
License
MIT