Every WaveSpeed model from your terminal.
One open-source CLI to search models, inspect schemas, estimate cost, run generations, upload inputs, and download outputs.
~/project $ wavespeed models z-imagescanning live model index...found wavespeed-ai/z-image/turbo~/project $ wavespeed schema wavespeed-ai/z-image/turbo --jsoninputs: prompt, size, seed, output_formatrequired: prompt~/project $ wavespeed price wavespeed-ai/z-image/turbo -p "a product photo"estimated cost: ready~/project $ wavespeed run wavespeed-ai/z-image/turbo -p "a clean product photo" --json --downloadcompleted ./wavespeed-output/0.pngOne command surface, every model family.
Instead of learning a different interface for each model type, use the same `models`, `schema`, `price`, `run`, `upload`, and `download` commands everywhere.
Text to image
z-image, Nano Banana, FLUX
Image to image
edit, enhance, remix
Text to video
Wan, Kling, Seedance
Image to video
animate a source image
Audio
speech, music, sound
3D
assets and generation tools
The workflow to remember.
Search a model, inspect its live inputs, estimate price, run with JSON, then download files or inspect history.
Search models by name, type, category, or popularity.
Inspect required inputs before running a model.
Estimate cost before generating.
Run generations with JSON output for scripts and agents.
Save generated files locally.
Core commands, small enough to memorize.
The CLI keeps discovery, execution, files, cost, and history in one predictable tool.
Discover models
Search image, video, audio, 3D, LoRA, and popular models from one CLI.
Inspect inputs
Read model schemas before writing prompts or automation scripts.
Run generations
Use one command pattern across text-to-image, image-to-video, editing, audio, and more.
Upload inputs
Upload local images or media, then pass the returned CDN URL into a model.
Download outputs
Save remote URLs or generation outputs into predictable local folders.
Check cost and history
Estimate price, inspect balance, and revisit completed predictions.
Built for AI agents.
Codex, Claude Code, Cursor, and other coding agents can discover models, inspect schemas, run generations, and parse JSON outputs directly from the terminal.
wavespeed skill install
wavespeed models --type text-to-image --popular
wavespeed schema <model_id>
wavespeed run <model_id> -p "your prompt" --json --downloadScript it like any other command.
Use `execFile`, parse stdout, and keep generated URLs or downloaded file paths in your own workflow.
import { execFile } from "node:child_process";
import { promisify } from "node:util";
const execFileAsync = promisify(execFile);
const { stdout } = await execFileAsync("wavespeed", [
"run",
"wavespeed-ai/z-image/turbo",
"-p",
"a clean product photo",
"--json",
"--download"
]);
const result = JSON.parse(stdout);
console.log(result.outputs);
console.log(result.saved);Project defaults stay with the repo. Secrets stay local.
Keep shared defaults in `wavespeed.json`. Keep API keys in local CLI config or environment secrets.
Initialize
Start with a shared `wavespeed.json`, then run without repeating the model ID.
wavespeed init -y
wavespeed run -p "a product photo" --json --downloadAlias
Package repeatable model settings behind a short command name.
{
"defaultModel": "google/nano-banana-2/text-to-image",
"outputDir": "wavespeed-output",
"aliases": {
"hero": {
"model": "wavespeed-ai/z-image/turbo",
"input": {
"size": "1024*1024",
"output_format": "png"
}
}
}
}Frequently Asked Questions
WaveSpeed CLI is the wavespeed command line tool for running WaveSpeed models from a terminal, script, CI job, or AI coding agent.
The same workflow can call text-to-image, image-to-image, text-to-video, image-to-video, audio, 3D, LoRA, and other WaveSpeed models.
--json gives Codex, Claude Code, Cursor, scripts, and CI jobs structured output they can parse reliably.
Keep API keys in local CLI configuration or environment-specific secrets. Do not commit keys to git, README files, or frontend code.