自主AI视频工程师

🦁 RemoQwen-MCP v8.0 • ETERNAL WATCHER

✅ 100% FREE & OPEN SOURCE
No API Keys • No Subscriptions • No Cloud • 100% Local Forever

The Most Powerful Autonomous AI Video Engineer for Remotion
Connects Qwen Desktop (Local AI) + Telegram Remote Control + Infinite Persistent Loop + Sentinel Lion
NEW: 🦁 Visual Self‑Healing via Browser Automation – AI sees your video, finds overlaps & sizing bugs, and fixes them autonomously.

✨ What is RemoQwen-MCP v8.0?

RemoQwen-MCP is a completely free local MCP server that transforms Qwen Desktop into a fully autonomous 24/7 Remotion motion graphics engineer.

You can send tasks from your phone via Telegram, close the laptop, and the AI will keep working non-stop.

Codename: Eternal Watcher (v8.0) — Stable, Immortal & Production Ready

🔥 Key Features

• 🦁 Sentinel Lion v3 – Concurrent multi‑point timeline crash detection (background, non‑blocking)
• 🌐 Visual Self‑Healing (NEW) – Opens Remotion Studio in a headed browser, inspects every keyframe for overlaps, misalignments, z‑index issues, text sizing, and other visual glitches; automatically fixes them in a closed loop.
• ♾️ Infinite Task Loop – wait_for_next_task() + reactive polling – AI never exits without permission
• 📱 Full Telegram Remote Control – Send prompts from phone, no terminal needed
• 🖼️ Asset Management – Upload images, list, preview, delete, and organise in public/ folder
• 🎥 Remote Rendering – Trigger video render via Telegram /render (permission‑aware)
• 📋 Message Utilities – /msgid to get message IDs, /del to delete messages (also cancels pending AI requests or clears tasks)
• 🔄 Background Verification – No MCP timeouts; AI polls status with verify_rendering_status
• 🧠 Persistent Memory – Top‑20 rules, auto‑saved on session end, loads all .md skills from .agents folder
• 🔐 Hybrid Permissions – Strict/Balanced modes require user approval (Telegram inline buttons + terminal)
• 🚦 3 Operation Modes – Fully Autonomous / Guarded Network / Strict Manual
• ❤️ Immortal SSE Connection – Heartbeat + 10‑hour keep‑alive
• 🎥 Cinematic Dashboard – Rich terminal with beautiful logs & icons
• 🌍 Zero Cloud Dependency – 100% local (Qwen Desktop + MCP server)
• 🔒 Security – Path jail, command whitelist, permission timeouts, SIGTERM handling, log level filtering

🚀 Quick Start

1. Create Remotion Project

   复制

   npx create-video@latest my-video
   cd my-video
   npm install
   

2. Clone & Setup

   复制

   git clone -b v8-remote-testing https://github.com/hse-solutions/RemoQwen-MCP.git
   cd RemoQwen-MCP
   pip install -r requirements.txt
   cp .env.example .env
   

3. Install Browser Automation Dependencies (NEW)

   复制

   playwright install chromium
   

   This downloads the Chromium browser used for visual inspection.

4. Configure .env (Very Important)

   复制

   # Required
   REMOTION_PROJECT_PATH=/full/absolute/path/to/your/my-video
   
   # Optional but recommended for full remote control
   TELEGRAM_TOKEN=1234567890:AAFxxxxxxxxxxxxxxxxxxxxxxxxxxxx
   AUTHORIZED_CHAT_ID=123456789
   
   # Optional environment variables
   LOG_LEVEL=INFO               # DEBUG, INFO, WARNING, ERROR
   PERMISSION_TIMEOUT=60        # seconds to wait for user approval
   DEEP_SCAN_POINTS=3           # number of verification frames (lower = faster)
   
   # Browser Automation (Visual Self‑Healing)
   REMOTION_STUDIO_PORT=3000    # port Remotion Studio runs on
   HEADLESS_BROWSER=false       # false = visible window (recommended)
   

5. Launch

   复制

   python run.py
   

   → Select mode → Copy the SSE link shown

📍 Qwen Desktop Setup

Step A: Create Folder & Add Infinite Loop Instructions

1. Open Qwen Desktop

2. Create New Folder → Name it RemoQwen-v8

3. Open the folder

4. Click ⚙️ Settings → Advanced Settings

5. In Custom Instructions paste this:

   复制

   when you start calling mcp you must read remote_task.md in src folder then you execute work user given from this file , after you finish your work you must need to run wait_for_next_task so if you have a answere from it STATUS_IDLE_NO_TASK , run again wait_for_next_task , so i mean while some work is coming to you from wait_for_next_task you need to run again and again this tool ok , you cant exit from this loop without user's authority
   

Step B: Connect MCP + SSE Link

1. Still in Advanced Settings → MCP Connections
2. Click + Add MCP
3. Fill:
   • Name: RemoQwen-MCP v8
   • Type: SSE
   • Server URL: http://127.0.0.1:8000/sse
   • Enable: ON
4. Click Save

📱 Telegram Remote Setup

1. Go to @BotFather → /newbot → Copy the TOKEN
2. Go to @myidbot → Send /getid → Copy your CHAT ID
3. Paste both in .env file (as shown above)
4. Restart python run.py
5. Now send any message to your bot → AI starts working instantly!

📱 Telegram Commands & Remote Control

Once your bot is running, you can control everything from Telegram. Send /help to see the available commands:

🚀 REMOTE COMMANDER v8.0

💬 Send Text - Starts a new mission
🛑 Send 'STOP' - Terminates the AI loop
🖼️ Image + Caption - Saves asset with custom name

📊 /assets - Choose folder (public/out) and list files
📷 /show_public <filename> or /showpub <filename> - View asset from public folder
🎬 /show_out <filename> or /showout <filename> - View rendered video from out folder
🎥 /render [composition] - Trigger video rendering (optional composition ID)
🗑️ /delete [name] - Remove asset from public folder
📋 /msgid [reply to a message] - Get message ID
🗑️ /deletemsg <id> or reply with /del - Delete a message (also cancels pending AI request or clears task)
📡 /status - Check system radar

How to use asset commands:

• Add an asset – Send an image with a caption. The caption will be used as the filename (e.g., logo.png). The image is saved inside your Remotion project's public/ folder.
• List assets – Type /assets → choose PUBLIC FOLDER (for images) or OUT FOLDER (for rendered videos).
• Preview asset – /show_public logo.png sends the image directly to Telegram. For videos: /show_out video.mp4.
• Delete an asset – Type /delete filename (e.g., /delete logo.png). You'll get a confirmation button.

Managing messages & pending requests

• Get message ID – Reply to any message with /msgid → bot replies with the message ID.
• Delete a message – Reply with /del (or /deletemsg) to delete that message. If the message was a permission request, the pending AI action is cancelled. If it was a text prompt or /render command, the pending task is cleared from remote_task.md.

🌐 Visual Self‑Healing (NEW – Browser Automation)

The AI can now open a headed Chromium browser and inspect the Remotion Studio visually, catching problems that even Sentinel Lion cannot detect (overlaps, sizing, z‑index). It then fixes the code automatically and re‑inspects – a true closed‑loop quality assurance.

How It Works

1. AI renders the video and passes Sentinel Lion runtime checks.
2. AI opens Remotion Studio in the browser (open_remotion_studio).
3. AI jumps to key frames (navigate_to_frame) – start, middle, end, and any animation change points.
4. AI extracts DOM layout (get_dom_layout) – gets exact pixel positions and sizes of all elements, automatically detects overlaps.
5. AI checks console errors (get_console_errors) – picks up z‑index warnings, missing refs, etc.
6. AI takes screenshots (capture_screenshot) if it needs to confirm visual output.
7. If problems found → AI calls write_file to fix the code → re‑renders → inspects again.
8. Loop continues until the video is visually perfect.
9. AI closes the browser (close_browser) and renders the final production‑ready MP4.

Tools Available to the AI

These tools give the AI full access to Playwright's browser automation capabilities – the AI decides which tool to use based on the problem it needs to solve.

🧠 Background Verification (No Timeouts)

The AI uses two tools for timeline verification:

1. verify_rendering_start – Starts background verification, returns a task_id immediately.
2. verify_rendering_status – Poll with the task_id to get the result.

This avoids MCP timeouts and lets the AI work on other things while verification runs.

📂 Full Folder Structure

RemoQwen-MCP/
├── .env.example
├── .gitignore
├── LICENSE
├── README.md
├── config.py
├── requirements.txt
├── run.py
├── src/
│   ├── server.py                  ← MCP server (background verification, tools)
│   ├── ui/
│   │   └── dashboard.py           ← Cinematic terminal UI
│   └── tools/
│       ├── asset_ops.py
│       ├── browser_ops.py         ← NEW: Browser automation & visual self‑healing
│       ├── file_ops.py
│       ├── memory_ops.py
│       ├── remote_ops.py          ← Telegram bot (all commands)
│       └── shell_ops.py
└── .env

Skills & Memory:

• Place your Remotion best practices / guidelines as .md files in the .agents folder inside your Remotion project.
• The AI will load all .md files recursively from .agents on startup.
• Memory (top 20 rules) is stored in memory.md at the project root.

🎯 Master Prompt Example

Initialize context. Create a premium 8-second liquid glass animation (1080p, cinematic). After writing the code, trigger Sentinel Lion verification, then open the studio, visually inspect frames 0, 100, 200, and fix any overlaps or sizing issues before final render.

🛠️ Advanced Configuration

You can set these in .env to fine‑tune the system:

🔒 Security Highlights

• Path jail – All file operations stay inside your Remotion project root.
• Command whitelist – Only npm, npx, node, remotion allowed.
• Hybrid permissions – In Strict/Balanced modes, sensitive tools require user approval (Telegram or terminal).
• Process timeout – Long‑running commands are automatically killed after 10 minutes.
• Telegram authorization – Only the chat ID you set in .env can interact with the bot.
• SIGTERM handling – Graceful shutdown on system signals.

🎁 Bonus Tip – AI Voiceover with Edge TTS Universal (Free, No GPU, No API Key)

Want to add high-quality AI voiceovers to your Remotion motion graphics videos? Use edge-tts-universal — a TypeScript library that uses Microsoft Edge's online TTS service with hundreds of natural Neural voices across 100+ languages. No GPU required, no API key, no model downloads, completely free.

Edge TTS Universal Repository: https://github.com/travisvn/edge-tts-universal
Full Features Reference (give this link to AI): https://github.com/travisvn/edge-tts-universal/blob/main/README.md

The Fully Autonomous Voiceover + Motion Graphic Workflow

Here is the exact procedure to make the AI automatically generate a voiceover and build a complete motion graphic video — all in one prompt:

Step 1 — Install the package manually (run this once in your Remotion project folder):

npm install edge-tts-universal

That's it. One package. No GPU drivers, no CUDA, no model downloads. It works immediately.

Step 2 — Give the Edge TTS Universal README to the AI so it understands how to use the library. Copy this link and paste it into your chat with Qwen Desktop:

https://github.com/travisvn/edge-tts-universal/blob/main/README.md

The AI will read the full API reference — all voices, prosody options, streaming API, subtitle generation — and know exactly how to write the voiceover script.

Step 3 — Send your prompt (via Qwen Desktop or Telegram). The AI will:

1. Create generate-voice.js using edge-tts-universal
2. Run node generate-voice.js to produce the voiceover MP3
3. Build a Remotion composition with <Audio> synced to the narration
4. Run Sentinel Lion verification + visual self-healing
5. Render the final video — fully autonomously

Example: generate-voice.js

The AI will create a file like this in your Remotion project root:

import { EdgeTTS } from 'edge-tts-universal';
import { writeFileSync } from 'fs';

// Initialize Edge TTS with a natural Neural voice
const tts = new EdgeTTS(
  'Welcome to Python. Python is one of the most popular programming languages in the world, known for its simplicity and power.',
  'en-US-EmmaMultilingualNeural',
  {
    rate: '+0%',     // Adjust speaking speed: +20% faster, -10% slower
    volume: '+0%',   // Adjust volume: +50% louder, -20% quieter
    pitch: '+0Hz',   // Adjust pitch: +5Hz higher, -5Hz lower
  }
);

// Generate the voiceover
const result = await tts.synthesize();

// Save as MP3 to public folder (accessible in Remotion)
const audioBuffer = Buffer.from(await result.audio.arrayBuffer());
writeFileSync('public/voiceover.mp3', audioBuffer);
console.log('✅ Voiceover saved to public/voiceover.mp3');

Run It

node generate-voice.js

The generated public/voiceover.mp3 is now accessible inside your Remotion compositions via <Audio src={staticFile('voiceover.mp3')} />.

Prompt to Give AI for Full Voiceover + Motion Graphic Video

Copy and paste this prompt to Qwen Desktop (or send via Telegram):

First, read and understand the Edge TTS Universal features from this link: https://github.com/travisvn/edge-tts-universal/blob/main/README.md

Then, generate a high-quality AI voiceover and create a motion graphics video:

1. Create a file called generate-voice.js that uses edge-tts-universal (import { EdgeTTS } from 'edge-tts-universal') with the voice 'en-US-EmmaMultilingualNeural'. The narration text should be: "Your narration script here". Save the output MP3 to public/voiceover.mp3.

2. Run 'node generate-voice.js' to generate the voiceover.

3. After the voiceover is ready, create a Remotion composition that plays the voiceover audio from public/voiceover.mp3 using <Audio src={staticFile('voiceover.mp3')} /> and synchronizes motion graphics animations to match the narration timing. Use <Sequence> components to time visual elements with the audio.

4. After writing the code, run Sentinel Lion verification, then visually inspect the studio and fix any issues before rendering the final video.

Change the narration text inside the prompt to match your video script. The AI will handle everything from voiceover generation to final render — fully autonomous.

Popular Voices

To discover all available voices, the AI can use:

import { VoicesManager } from 'edge-tts-universal';

const voicesManager = await VoicesManager.create();
const englishVoices = voicesManager.find({ Language: 'en' });
const femaleUSVoices = voicesManager.find({ Gender: 'Female', Locale: 'en-US' });
console.log(englishVoices.map(v => v.ShortName));

Why Edge TTS Universal Over Other TTS Options?

MIT License • Completely Free & Open Source Forever
Made with ❤️ by HIRUNA