Transcript Refine

English | 中文

Obsidian plugin to refine speech-to-text content with AI. Fix errors, improve structure, and turn messy transcripts into clear, readable text.

Features

• Refine selection: Select text → Command Palette or right-click → AI refine → replaces in place
• Refine whole document: One-click to restructure the entire document
• Four preset templates:
  • General — daily voice notes, fix typos and grammar, split into paragraphs
  • Meeting Minutes — extract topics, group by agenda, list discussion points and action items
  • Interview — preserve Q&A structure, trim redundancy, highlight key quotes, append summary
  • Polish — preserve spoken style, make expression smoother and more natural
• Custom templates: Add, edit, and delete your own refine templates
• Right-click menu: Select text, right-click to choose a template
• Multi-API support: Optimized for DeepSeek, compatible with OpenAI and all compatible APIs
• Secret storage: API key managed via Obsidian SecretStorage — never written to plaintext config files

Usage

1. Fill in API URL and Model in settings, select your API Key from the system keyring
2. Choose a default template
3. Select the speech-to-text content in the editor
4. Trigger via:
   • Command Palette (Ctrl+P) → "AI 整理选中文字" or "AI 整理整篇文档"
   • Right-click → "AI 整理" or "AI 整理为 →"
5. The refined text replaces the original in place — undo with Ctrl+Z if unhappy

Settings

Setting	Description	Default
API URL	DeepSeek-compatible API endpoint	https://api.deepseek.com
API Key	Select or create via system keyring (SecretStorage)	—
Model	Model name	deepseek-v4-flash
Default template	Template used for quick refine	General
Timeout	API timeout (ms)	30000
Max tokens	Output token limit	4096

Installation

Manual

1. Download main.js, manifest.json, styles.css
2. Place them in <Vault>/.obsidian/plugins/transcript-refine/
3. Enable the plugin in Obsidian settings

Development

npm install
npm run dev    # Watch mode
npm run build  # Production build