Documentation
Set up CueTwin your way.
CueTwin ships with no accounts, keys or voices of its own. Everything below uses your own logins, your own recordings and your own keys, and they stay on your PC.
Install CueTwin
- Download
CueTwin-Beta-Setup-x64.exe(Windows 10/11, 64-bit). - Run it. The installer is not code-signed yet, so Windows SmartScreen may show “Windows protected your PC”. Choose More info, then Run anyway.
- CueTwin installs for your Windows user only (no administrator rights). You can change the install folder, and it adds Desktop and Start menu shortcuts.
Windows 10/11 64-bit · v1.2.0 · AI models (roughly 8 GB) download on first launch. View releases on GitHub
Each release lists the installer’s SHA-512 in latest.yml on the GitHub release page.
First launch and AI models
The installer does not include the AI models. On first launch CueTwin downloads them (roughly 8 GB), checks every file, and picks Qwen3.5 9B (“Pro”) when your GPU has room or Qwen3.5 4B (“Lite”) otherwise. You can revisit this any time from Help › AI Models.
Free during the public beta. Paid licences come later through Lemon Squeezy. During the beta every feature is unlocked, with no trial or time limit.
Updates
- CueTwin checks for a new release about 15 seconds after it starts and then every six hours.
- A newer version downloads in the background and is verified (size and SHA-512) before anything is offered.
- When it is ready, an update banner appears. Choose Install update & restart. CueTwin never restarts by itself, and your projects, voices and settings are kept.
- To check right away: Help › Check for Updates…
The check downloads one small file (latest.yml) from GitHub Releases. It sends no account, project or usage data.
Connect an AI assistant
Open Settings and choose a provider. The built-in assistant works with nothing else installed. The others use your own account or install:
| Provider | What you need |
|---|---|
| CueTwin (built-in) | Runs locally with the models CueTwin downloads. No account, no key. |
| Claude | Either the Claude Code CLI with your own Claude login, or your own Anthropic API key. |
| Cursor | The Cursor CLI, signed in with your own account. |
| Local (Ollama) | Your own Ollama install, default address http://127.0.0.1:11434. |
| OpenAI, OpenRouter, Groq, Grok (xAI), DeepSeek | Your own API key for that service. |
| Custom | Any OpenAI-compatible endpoint: set the Base URL and, if needed, a key. |
Cloud providers receive your prompts and bill you directly. API keys you enter are kept on this PC.
Clone your own voice
Voice cloning runs locally. CueTwin ships with no voices: you add recordings of a voice you are allowed to use, normally your own.
- Instant clone: add a few clean samples and CueTwin picks the best reference. No training needed.
- Professional clone: add about ten minutes of usable speech. CueTwin trains local candidates and keeps a voice only if it passes checks for identity, wording, pacing and noise.
- The first time you use voice features, CueTwin sets up its speech runtime. It needs Python 3.11 (64-bit) from python.org and an internet connection to fetch the speech and transcription models.
Recordings and trained voices never leave your PC and are never part of a CueTwin download or update.
ElevenLabs (optional)
If you have an ElevenLabs account you can use its voices alongside local ones. Paste your own API key in the ElevenLabs settings. CueTwin stores it encrypted with Windows secure storage, tied to your Windows account. Text you send to ElevenLabs is processed and billed by ElevenLabs.
Control CueTwin from Claude Code, Codex and other agents
CueTwin exposes its editor tools through a local, token-protected API and an MCP server. An agent can inspect the timeline, trim, split, add text and transitions, generate voice, import media, save and export, without touching your mouse or keyboard. Keep CueTwin open.
The MCP adapter needs Node.js 20 or newer. With the default install folder:
claude mcp add -s user cuetwin -- node "%LOCALAPPDATA%\Programs\CueTwin\resources\automation\automation-mcp.cjs"
codex mcp add cuetwin -- node "%LOCALAPPDATA%\Programs\CueTwin\resources\automation\automation-mcp.cjs"In PowerShell use $env:LOCALAPPDATA. Tools: cuetwin_state, cuetwin_tools, cuetwin_call, cuetwin_job, cuetwin_cancel.
- The service listens only on 127.0.0.1, needs the bearer token in
%APPDATA%\CueTwin\automation.json, rejects requests from web pages, and offers no shell or arbitrary code execution. - Every change goes through CueTwin’s normal undo history. Work on another timeline or project happens in a hidden editor and never moves what you are looking at.
Where your data lives
- Projects are folders you choose; media, recordings, narration and exports stay inside.
- Settings, voices, keys and models live in
%APPDATA%\CueTwin(or the data folder you picked). - The installer contains none of the developer’s data: no voices, recordings, projects, API keys, logins or Claude connection. Every user starts empty and connects their own.
Uninstall
Use Windows Settings › Apps. Uninstalling keeps %APPDATA%\CueTwin so your voices and settings survive a reinstall. Delete that folder yourself to remove them.