vercel-labs/ai-cli

What it solves

This project provides a unified, command-line interface (CLI) for interacting with a wide variety of AI models across different modalities. It eliminates the need to write custom scripts for simple tasks like generating images, video, audio, and text, and simplifies the process of chaining multiple AI operations together using standard terminal piping.

How it works

Built on the Vercel AI SDK and AI Gateway, the CLI allows users to execute simple commands (e.g., ai text, ai image, ai video, ai audio) to call various models. It supports stdin for input, meaning you can pipe data from other terminal commands (like git diff or cat) directly into an AI model. It also handles the output of one AI generation to be the input for another, allowing for complex workflows like generating an image and then animating it into a video.

Who it’s for

Developers and power users who prefer the terminal over a web UI, and those who want to quickly experiment with different models from different providers without managing multiple API keys or separate toolsets.

Highlights

• Multi-modal support: Generates text, images, video, and audio (speech and transcription).
• Piping and references: Supports stdin and reference images/files for complex, agent-native workflows.
• Unified model access: Accesses hundreds of models via AI Gateway with a single interface.
• Inline previews: Displays images, videos, and audio waveforms directly in supported terminals (e.g., Kitty, Ghostty, WezTerm).
• Model comparison: Ability to generate outputs from multiple models simultaneously for comparison.
• Detailed model info: Provides live data on pricing, latency, and throughput for available models.