ReaperAI

# ReaperAI v1.0.3 ReaperAI is an AI assistant that runs within REAPER, converting natural language instructions into confirmable and executable REAPER operations. It helps audio engineers, music producers, and game audio designers complete tasks such as track organization, media processing, Region/Marker management, batch operations, and audio generation. You can describe your needs like this: - Create 10 footstep sound tracks and name them by material - Add 80ms fade in and out to all selected media - Delete Regions 5 to 10 - Rename all Regions to SFX_001 format - Play or stop the project - Generate a mechanical door opening sound effect and import it into the project ReaperAI will generate an operation plan, which you can confirm before execution. By default, it won't modify the project directly. ## Demo ### Normal Chat  ### Execution Mode: Create Tracks  ## Features - Natural Language Control: Describe operations in English or Chinese without needing to remember Action IDs or ReaScript APIs. - Confirmation Before Execution: AI generates an operation card, and you confirm before execution. - MCP Enhanced Execution: Connect to a local MCP server for more stable structured capability endpoints. - MCP Offline Fallback: If MCP fails to start, common operations will use local Lua/SCRIPT methods. - Local Capability Detection: Detects available APIs, Action IDs, and extensions in the current REAPER environment. - LLM Vendor Configuration: Supports multiple OpenAI-compatible APIs and refreshes model lists. - ElevenLabs Integration: Optional voice and sound effect generation capabilities. - Risk Alerts: Batch modifications, file writes, and deletions are marked as higher-risk operations. - Local Execution: Main logic runs in the local REAPER, Lua, and Python environments. ## Supported LLM Vendors You can choose from the settings page: - OpenAI - DeepSeek - Moonshot / Kimi - MiniMax - Qwen / DashScope - SiliconFlow - OpenRouter - Custom OpenAI-Compatible After selecting a vendor, ReaperAI will automatically fill in the recommended Base URL, default model, and model list. The model list is updated frequently, so the settings page also provides a "Refresh Model List" feature to fetch the latest models from the vendor's `/models` interface. ## System Requirements - Windows 10 / Windows 11 - REAPER 7.x - A valid LLM API Key - Network access to your chosen LLM service No prior Python installation is required. The project includes an installation script that prepares a private Python environment for ReaperAI and installs core dependencies offline using wheels. ## Project Structure ```text ReaperV1.0/ ├── MCP_Server/ Local MCP service, LLM worker, and installation scripts ├── Scripts/ │ ├── ReaperAI.lua REAPER entry script │ ├── rai_async_pipe.lua Asynchronous request pipe │ └── ReaperAI/ ReaperAI core module ├── UserPlugins/ ReaImGui, js_ReaScriptAPI, ReaPack, SWS, etc. ├── README.md ├── LICENSE └── installation_guide.txt ``` ## Installation 1. Download the project. 2. Copy the `MCP_Server`, `Scripts`, and `UserPlugins` folders to your REAPER root directory, e.g., `X:\REAPER`. 3. In the `MCP_Server` folder, double-click `【Step 1】Install Dependencies.bat`. 4. Double-click `【Step 2】Configuration Wizard.bat` and follow the prompts to complete the configuration. 5. Open REAPER, press `?` to open the Action List. 6. Click `Load`, select `X:\REAPER\Scripts\ReaperAI.lua`. 7. Run `ReaperAI.lua`. If your REAPER is not in `X:\REAPER`, choose the correct path for the script. ## Initial Configuration After opening ReaperAI, go to the "Settings" tab: 1. Choose a provider in "LLM Configuration". 2. Enter the corresponding API Key. 3. Select or input a Model Name. 4. Click "Test Connection" to confirm the API is available. 5. Click "Refresh Model List" to update the available models. 6. In the "Local Capability Detection" area, click "Detect". 7. Save the configuration. You can then switch to the "Conversation" tab to use ReaperAI. ## Usage ### Consultation Mode By default, ReaperAI only answers questions and doesn't execute any REAPER operations. Suitable for asking: - How to perform an operation? - Where is a specific REAPER feature? - Is a mixing or sound design approach reasonable? ### Execution Mode After clicking "Execution Mode", ReaperAI enters an executable state. You can send natural language instructions, and AI will generate an operation plan for confirmation. Execution mode doesn't strongly depend on MCP: - MCP successfully started: Uses MCP enhanced capabilities. - MCP fails to start or times out: Retains local execution mode, using Lua/SCRIPT for common operations. - Only clicking "Exit Execution" will leave execution mode. ### Audio Generation If you have configured an ElevenLabs API Key, you can use the "Generate Audio" tab to create voice or sound effects and import them into your REAPER project. ElevenLabs is an optional feature; not configuring it won't affect core conversations and REAPER operations. ## Executable Operations ReaperAI currently supports these tasks: - Play, stop, and other transport controls - Create, delete, rename tracks - Set track volume, pan, mute, solo - Add, delete Markers - Add, delete, batch rename Regions - Process selected media's fades - Batch organize media or tracks - Read project context and generate operation suggestions - Generate voice or sound effects via ElevenLabs More complex or high-risk operations require confirmation, and insufficient information will prompt clarification questions. ## MCP vs. Local Execution ReaperAI has two execution capabilities: | Mode | Description | | --- | --- | | MCP Enhanced Execution | Used when the local Python MCP service is online, suitable for structured and verifiable REAPER operations | | Local Execution Mode | Used when MCP is offline, falls back to Lua/SCRIPT for common operations | Even if issues arise with the user's Python, network, or MCP service, ReaperAI won't directly lose execution mode. However, some advanced capabilities, like audio analysis, complex export, or sound effect generation, may still require MCP or external APIs to function. ## Security Mechanisms ReaperAI explicitly presents execution risks: - Non-execution mode won't output executable operations. - Execution mode generates an operation plan, not directly automating the project. - Delete, batch modify, and file write operations display risk alerts. - AI-generated scripts are processed by a local executor. - Local capability detection generates a list of available APIs and Actions, reducing the likelihood of model calls to non-existent APIs. It's recommended to save your project before using ReaperAI for complex operations or test them in a project copy. ## Troubleshooting ### Connection Test Timeout Please check: - API Key correctness - Provider and Base URL match - Current network access to the corresponding vendor - Proxy or firewall settings - Model name existence If "Refresh Model List" fails but local recommended models are available, you can use the built-in model list. ### MCP Startup Failure After Clicking Execution Mode This is an acceptable degraded state. ReaperAI retains local execution mode, and common operations are still executable. If you need MCP enhanced capabilities, check: - Whether `【Step 1】Install Dependencies.bat` was run - If `MCP_Server` is in the REAPER root directory - Port `8765` availability - Antivirus software not blocking local Python or bat scripts ### Local Capability Detection Failure Please confirm: - `Scripts` folder copied to REAPER root directory - `UserPlugins` folder copied to REAPER root directory - REAPER restarted - Current user has permission to write JSON files in the REAPER directory After successful detection, ReaperAI uses the generated capability list to improve execution compatibility. ## Privacy Statement ReaperAI sends your conversation content, current project context, and necessary operation intent to your configured LLM service provider to generate responses or operation plans. API Keys are stored in local configurations; please don't upload your configuration files or keys to public repositories. The project itself doesn't embed any user's LLM API Key. ## License This project uses the MIT License. See [LICENSE](LICENSE). ## Author zhanghuisen GitHub: [Zhanghuisen](https://github.com/Zhanghuisen)