Mimo TTS Markdown Reader

Read aloud editor documents using the MiMo LLM Text-to-Speech API (v2.5). Supports Markdown and plain text with Director Mode for fine-grained voice control.

Features

• 📖 One-click reading: Command palette MiMo: Read Current Document or right-click in editor
• ✂️ Selection reading: Read only the selected text, or the entire document if nothing is selected
• ⏯️ Full playback control: Start / Pause / Resume / Stop at any time
• 🎬 Director Mode: Auto-inject style tags with semantic enhancement for Markdown structures
• 📊 Status bar indicator: Real-time display of playing (with progress) / paused / stopped
• 📑 Long document splitting: Automatically split by sentence boundaries, synthesize and play in order
• 🔌 Extensible architecture: TTS service abstracted as an interface for easy integration with other providers

Getting Started

1. Configure API Key

Open VS Code Settings, search for mimoTts, and fill in:

2. Commands

You can also right-click in the editor and select "Read This Document".

Architecture

src/
├── extension.ts              # Entry point: activate/deactivate, command registration, config listener
├── types.ts                  # Type definitions: ITtsService interface, PlaybackState, TtsConfig
├── configManager.ts          # Configuration reader (supports hot-reload)
├── preprocessor.ts           # Director Mode text preprocessing: Markdown semantic recognition + style tag injection + segmentation
├── ttsClient.ts              # MiMo TTS v2.5 API client (ITtsService implementation)
├── audioPlayer.ts            # Audio player: temp files + system player
├── playbackController.ts     # Playback controller: state machine + segment queue + cursor management
├── statusBar.ts              # Status bar indicator
└── test/                     # Unit tests
    ├── preprocessor.test.ts
    ├── ttsClient.test.ts
    └── runTest.ts

Module Dependencies

extension.ts (entry/commands)
    │
    ▼
PlaybackController (state machine + queue + status bar)
    │            │
    ▼            ▼
preprocessor   MiMoTtsClient (ITtsService)
(text→tags)        (HTTP requests)
                     │
                     ▼
               AudioPlayer (temp files + system player)

Design Highlights

• TTS Service Abstraction: The ITtsService interface isolates vendor differences. The controller layer is independent of any specific implementation. Adding other TTS providers only requires a new implementation class.
• State Machine: idle → playing → paused → playing → stopped. Pause uses "stop current segment + record cursor" semantics.
• Segmentation Strategy: Split by sentence endings/newlines, 1–4 sentences per segment, aligned with MiMo Director Mode guidelines.

Director Mode

MiMo TTS v2.5 Director Mode Specification

• Style tags are placed at the beginning of the text in the format: (style1 style2)content to synthesize
• Style tags are placed in the messages[role=assistant].content field
• Each segment should contain 1–4 sentences for clear semantics
• Supported delimiters: half-width (), full-width （）, square brackets []

Markdown Structure Recognition

The plugin recognizes Markdown structures line by line and automatically adds style tags:

Usage Example

Input Markdown

# MiMo TTS Plugin Guide

This plugin can read aloud document content in the current editor.

> Text-to-speech makes document proofreading more efficient.

- Supports full document reading
- Supports selection reading
- Supports pause and resume

After Director Mode Preprocessing (content in synthesis request)

Assuming config: defaultStyle = "narration", bracketStyle = "half"

Segment 1: (narration emphasis)MiMo TTS Plugin Guide

Segment 2: (narration)This plugin can read aloud document content in the current editor.

Segment 3: (narration calm quote)Text-to-speech makes document proofreading more efficient.

Segment 4: (narration pause)Supports full document reading

Segment 5: (narration pause)Supports selection reading

Segment 6: (narration pause)Supports pause and resume

Each segment is sent as an independent messages[role=assistant].content for synthesis, and audio is played in sequence.

Development & Build

# Install dependencies
npm install

# Compile TypeScript
npm run compile

# Build bundle (esbuild)
npm run package

# Debug (open this project in VS Code, press F5 to launch Extension Development Host)

Potential Improvements

1. Secure Key Storage: Currently the API Key is stored in VS Code Settings (plaintext). Can be migrated to vscode.SecretStorage for encrypted storage.
2. Streaming Playback: Currently uses batch requests + sequential playback. Can be changed to stream-and-play for lower first-segment latency.
3. Multi-platform Player: macOS (afplay) / Linux (xdg-open) branches are already prepared. Pause/stop control can be further improved.
4. Other TTS Providers: Implement the ITtsService interface to integrate other providers (e.g., Azure TTS, Alibaba Cloud TTS).
5. Progress Jump: Currently can only resume from the current segment after pausing. Can add "go back one segment / go forward one segment" controls.
6. Waveform Visualization: Integrate a Webview panel to display audio waveform and reading progress.
7. Internationalization: Add i18n support.