TubeChef CLI Setup Guide: Install, API Key and First Render
Set up TubeChefCLI on macOS (Apple Silicon) or Windows with an NVIDIA GPU: download from your dashboard, add your API key, prepare and render your first video.
Short answer: TubeChef renders videos on your own computer, not in the cloud. You can do that with the TubeChef desktop app (the easier option) or with TubeChefCLI, a command-line tool for automation and batch rendering. To set up the CLI:
- Download it from your TubeChef dashboard.
- Install the requirements for your system.
- Run it and paste your API key.
- Mark videos for processing in the dashboard, then let the CLI prepare and render them.
Before you start
| macOS | Windows | |
|---|---|---|
| Hardware | Apple Silicon (any M-series Mac). Intel Macs aren't supported | x64 PC with an NVIDIA GPU (RTX). The dashboard lists at least an Intel i5 9th gen (or AMD equivalent) and 8 GB RAM. Older pre-RTX cards (e.g. GTX 1060) don't work |
| Extra software | Homebrew, FFmpeg, Python 3.12.3, PyTorch, OpenAI Whisper | CUDA Toolkit, latest NVIDIA drivers, Chocolatey, FFmpeg, Python 3.12, PyTorch (CUDA build), OpenAI Whisper |
| Plan | VIP or Lifetime | VIP or Lifetime |
Linux isn't supported. You'll also need a TubeChef API key, from the API Keys section of your dashboard.
Step 1: Download TubeChefCLI from your dashboard
Log in at tubechef.ai, open Account → Desktop/CLI, choose your platform and click Download. You get a zip with the TubeChefCLI program. Unzip it into a folder you'll keep, e.g. ~/TubeChef or C:\TubeChef.
Only download TubeChef from your dashboard.
Step 2: Install the requirements
The dashboard page has the exact commands for each system, plus a video tutorial. In summary:
macOS (Apple Silicon)
- Install Homebrew, then
brew install ffmpeg. - Install Python 3.12.3 (the macOS 11+ installer linked on the dashboard).
pip3 install torch torchvision torchaudiopip3 install -U openai-whisper- Check everything:
ffmpeg -version,python3 --version,whisper --help.
If macOS blocks TubeChefCLI the first time ("cannot be opened because the developer cannot be verified"), go to System Settings → Privacy & Security and click Open Anyway for TubeChefCLI. The app isn't Apple-notarized yet, so this warning is expected.
Windows (NVIDIA GPU): run PowerShell as Administrator.
- Install the CUDA Toolkit and the latest NVIDIA driver.
- Install Chocolatey, then
choco install ffmpeg -yandchoco install python312 -y. - Install the CUDA build of PyTorch and Whisper. RTX 50-series cards need the nightly build; the dashboard shows both commands.
- Check everything:
nvcc --version,nvidia-smi,ffmpeg -version,python --version,whisper --help.
Step 3: Run it and add your API key
Open a terminal in the folder and run the program with no arguments:
./TubeChefCLI # macOS
TubeChefCLI.exe # Windows
The first time, it asks for your API key. It stores the key on your computer, and you can change it later under Settings → Change API key. For scripts or servers, you can set the environment variable TUBECHEF_API_KEY instead.
One key, one computer. Your API key is tied to the first device that uses it. If you try it on a second computer, you'll get "API key bound to a different device". To move to a new computer, contact support.
Step 4: Your first render
- In the web dashboard, create your video (script, voice, images) and mark it for processing.
- In TubeChefCLI, choose Prepare assets. It downloads the script, audio and images for the videos you marked into local folders. You can add your own files to those folders before rendering.
- Choose Render prepared assets. FFmpeg assembles each chapter and exports an MP4.
- Or choose Full pipeline (automatic) to prepare and render in one go. The auto mode checks for newly marked videos every 5 minutes, which is handy for leaving a machine rendering overnight.
To clean up disk space, choose Clean Videos to delete downloaded assets after rendering.
Commands for automation (recent versions)
Interactive mode covers most needs. Recent versions also accept commands, which the desktop app uses and which you can use in your own scripts:
| Command | What it does |
|---|---|
TubeChefCLI --version |
Shows the installed version |
TubeChefCLI diagnose |
Checks the bundled tools (FFmpeg, Whisper, etc.) |
TubeChefCLI prepare [--channel <id>] [--video <id>] |
Downloads assets for videos marked for processing |
TubeChefCLI channels |
Lists locally prepared channels and videos |
TubeChefCLI render <channelId> <videoId> |
Renders one prepared video |
TubeChefCLI srt |
Processes pending subtitle (SRT) jobs |
--json-output |
Machine-readable output for scripts |
--output-dir <path> |
Where rendered videos go |
Channel Ripper commands are available too (ripper-discover, ripper-rip).
Troubleshooting
- "No API key provided": run the program with no arguments to add a key, or set
TUBECHEF_API_KEY. - "API key bound to a different device": the key is tied to another computer; contact support.
- "No videos marked for processing": mark videos for processing in the dashboard first.
- Whisper/caption errors on Windows: check
nvidia-smiand that PyTorch reports "CUDA available: True". Pre-RTX GPUs aren't supported. - Something's missing: run
TubeChefCLI diagnose.
FAQ
Is TubeChef CLI on npm?
No. Download TubeChefCLI from your dashboard. npm install tubechef-cli isn't a TubeChef package.
Does it work on Linux or an Intel Mac?
No. Only macOS on Apple Silicon and Windows x64 with an NVIDIA GPU.
CLI or desktop app?
The desktop app bundles the same engine with a visual interface, and it's the easier start. Use the CLI for automation, servers and scripted batches.
Is there cloud rendering?
No. Rendering always runs on your own computer.