Skip to content

Install the Mac app

Reelfold for Mac is the desktop app: Home to start and follow your work, an Inbox for what needs you, All projects for every video, and assisted publishing. It is free and MIT licensed. This page gets it running and connected to an AI model.

  • A Mac with Apple Silicon (M1 or later). Windows is in preview: see Install on Windows.
  • Disk space for the app, the speech model (about 1.6 GB) and your renders.
  • An AI model you already have access to. Any one of these works: a Claude Code or Codex login, an API key, or a local model. You can also start with none; planning then falls back to simple rules.
  1. Open github.com/zyziyun/reelfold/releases/latest and download the .dmg for Apple Silicon (mac-arm64).
  2. Open the DMG and drag Reelfold into Applications.
  3. Start Reelfold from Applications.

The installed app is signed and notarized, and carries its own engine: Python, ffmpeg and the vstudio engine are bundled, so you do not need to install them yourself.

A Mac App Store edition (listed as Reelfold, still in Apple’s review) runs in Apple’s sandbox: it uses API keys or local models for AI (no Claude Code or Codex sign-in) and updates through the App Store. See Editions.

You need Node 22+, Python 3.10+ and ffmpeg (for example brew install node python ffmpeg).

  1. Clone the repository and install the engine and the app’s packages:

    Terminal window
    git clone https://github.com/zyziyun/reelfold && cd reelfold
    ./install.sh && npm install

    install.sh installs the Python dependencies and downloads fonts and face models into ~/.cache/video-studio.

  2. Start the app:

    Terminal window
    npm run desk

    The app opens as Reelfold and starts its engine in the background, using the lib/ folder of this checkout.

If the app picks the wrong Python, set the Python path in Settings.

When Reelfold starts, it launches a local engine on your Mac (a helper process the app talks to over a private local socket; it listens on no network port). The status shows Ready on this computer once it is up.

A short first-run setup walks through AI, downloads and your platforms (you can run it again from Settings › Advanced). The installed app downloads what the engine needs in the background; progress is under Settings › Advanced › Downloads:

Download Size Needed for
Fonts and MediaPipe face models about 62 MB Captions, covers, face tracking, retouch
Whisper large-v3-turbo about 1.6 GB Transcription
Free voice (Kokoro) about 330 MB Narration without an OpenAI key; downloaded when a video first needs it
Chromium (optional) 100–120 MB HTML covers, slides and HyperFrames-based workflows

Downloads resume if interrupted and are checked before they are installed. In a source build, install.sh has already fetched the fonts and face models, and the engine downloads a Whisper model on the first transcription if none is installed.

Open Settings › AI. It shows the AI that is working and when it was last checked; each provider shows a status (signed in, login expired, key saved, local server running).

Option What you do Billing
Subscriptions (no API key) Click Sign in next to Claude Code or Codex (install the CLI first). Your browser opens; if it shows a code, paste it into the app. Your Claude Pro / Max or ChatGPT subscription
API keys Paste a key for Anthropic, OpenAI, DeepSeek, Qwen, Kimi, GLM, OpenRouter, Gemini or ElevenLabs. Keys are kept in the macOS keychain and handed only to the engine. Pay per use
Local models Run Ollama, LM Studio, vLLM or llama.cpp on this Mac or your own server. Reelfold lists the models it finds. Free; nothing leaves your network

Use Test on a provider to send one tiny request and confirm it answers.

Change order sets which AI is tried next when the first doesn’t answer, and Different AI for specific jobs sets a provider per task (planning, picking segments, captions, post copy, AI edit). When a backup answers, the app tells you which one did. A new API key needs an engine restart; the page offers the button.

What each step sends to the provider, and what stays local: AI providers.

  • Installed app: Reelfold checks for updates at launch and every few hours, downloads them in the background and offers Restart to update (with what’s new). If projects are running, Restart when done waits for them. Settings › General shows the version and Check for updates.

  • App Store edition: updates come from the Mac App Store.

  • Source build: pull and reinstall, then start again:

    Terminal window
    git pull && ./install.sh && npm install
    npm run desk

Drag Reelfold from Applications to the Trash. Your projects and settings stay on disk until you remove them:

Folder What is in it
~/Library/Application Support/Reelfold App settings, downloaded models and fonts (assets), encrypted API keys, the publish browser’s platform logins
~/.config/vstudio The engine’s home: the project list, publish calendar and your filler-cut preferences; new projects go to projects/ here by default
~/.cache/video-studio Engine cache: fonts, models, transcripts. Shared with the Claude Code skill
~/.cache/huggingface/hub Whisper models downloaded by a source build or the skill

Your original footage is never moved or modified. If you came from an older Daycut build, its profile folder is left in place as a backup the first time Reelfold copies it.