OpenArt CLI

Make images and video with OpenArt, from your terminal.

openart puts your OpenArt account on the command line. Describe what you want, pick a model, and the result comes back in the same window — as a link you can share, or a file saved straight to your computer.

It is one file to install and nothing else to set up: no Python, no Node, no API keys to copy and paste. Sign in once through your browser and you are ready.

openart login
openart generate image "a red fox in the snow" --model nano-banana-2
openart generate video "a paper boat drifting down a rain gutter" --model kling-3-omni

What you can do

• Make an image from a description. Give it a sentence and a model, and get a picture back.
• Change an image you already have. Point it at a photo on your computer or a link, and describe the edit — make it snowy, restyle it, move it somewhere else entirely.
• Make a video from a description, or bring a still photo to life. Where the model allows it, you choose the length, the shape and the resolution.
• See what anything costs before you spend a credit. List every model available to you, compare prices cheapest-first, and check exactly which settings a particular model accepts.
• Keep the results. Print a link, or download straight into a folder.
• Look back over everything you have made. Check on a job that is still running, or wait for it to finish.
• Stay organised. Switch between projects and workspaces, and upload reference images once to reuse them later.
• Check your plan and credit balance whenever you like.
• Let a program drive it. Every command can return plain data instead of a table, so scripts, pipelines and AI agents can use it as easily as you can.

Model names change as new ones arrive — run openart model list to see everything available to you today.

Contents

• Install — macOS and Linux · Windows · Manual download · Upgrading
• Getting started — signing in, and checking your credits
• Generating images and video — text to image, image to image, text to video, animating a photo
• Finding models, and what they cost — what is available, what it accepts, what it charges
• History, projects, and uploads — past results, organising work, reference images
• Scripting and agents — plain data out, and a safe way to rehearse a request
• Uninstall
• Support · License

Install

macOS and Linux

curl -fsSL https://raw.githubusercontent.com/OpenArt-AI/cli/main/install.sh | sh

The installer picks the right build for your machine, verifies it against the release checksums, and installs to /usr/local/bin — or to ~/.local/bin if that is not writable, rather than prompting you for sudo unexpectedly.

# install somewhere specific
curl -fsSL .../install.sh | sh -s -- --prefix "$HOME/.local"

# pin an exact version
curl -fsSL .../install.sh | sh -s -- --version 0.1.0

Windows

irm https://raw.githubusercontent.com/OpenArt-AI/cli/main/install.ps1 | iex

Installs to %LOCALAPPDATA%\Programs\openart\bin and adds it to your user PATH — no administrator rights required. A piped script cannot take parameters, so to pin a version set an environment variable first:

$env:OPENART_VERSION = '0.1.0'
irm https://raw.githubusercontent.com/OpenArt-AI/cli/main/install.ps1 | iex

Manual download

Grab the archive for your platform from the latest release, extract it, and put openart somewhere on your PATH.

Platform	Asset
macOS, Apple silicon	openart_<version>_darwin_arm64.tar.gz
macOS, Intel	openart_<version>_darwin_amd64.tar.gz
Linux, x86-64	openart_<version>_linux_amd64.tar.gz
Linux, arm64	openart_<version>_linux_arm64.tar.gz
Windows, x86-64	openart_<version>_windows_amd64.zip
Windows, arm64	openart_<version>_windows_arm64.zip

Every release ships a checksums.txt. Verify your download with:

shasum -a 256 -c checksums.txt --ignore-missing

Upgrading

Re-run the installer — it overwrites the existing binary in place.

Getting started

openart login       # opens your browser; no API key to copy and paste
openart account     # confirms who you are, your plan, and your credit balance

login uses the standard OAuth browser flow with PKCE. The credential is stored at ~/.openart/cli-credentials.json (mode 0600) — or %USERPROFILE%\.openart\cli-credentials.json on Windows — and refreshed automatically. openart logout revokes it and clears local state.

Generating images and video

Media is always explicit in the command — generate image or generate video — so an image request can never be routed to a pricier video model by accident.

# text to image
openart generate image "a red fox in the snow" --model nano-banana-2

# image to image: a local file, or a URL already on https://cdn.openart.ai
openart generate image "make it snowy" --model nano-banana-2 --image ./fox.png

# download the result instead of just printing its URL
openart generate image "neon city" --model gpt-image-2 -o ./out/

# text to video, and image to video
openart generate video "timelapse of a blooming flower" --model kling-3-omni \
  --duration 8 --aspect-ratio 16:9 --resolution 720p
openart generate video "slow zoom in" --model kling-3-omni --image ./photo.jpg

Each command submits the job, polls until it finishes (bounded by --timeout, default 5 minutes), and prints the result. Pass --async to skip the wait: the command submits, prints the generation's id, and exits, leaving you to pick the result up later with openart creation wait <id>. --duration, --aspect-ratio, and --resolution are model-specific — omit them to take the model's own defaults.

Finding models, and what they cost

openart model list                          # every model, and the modes it supports
openart model cost                          # price everything, cheapest first
openart model cost --model <id> --mode text2image
openart model form <id> text2image          # exactly which parameters that model takes

model cost never spends credits — it only quotes them.

History, projects, and uploads

openart creation list --type video --limit 10
openart creation get <id>                   # status and result URLs
openart creation wait <id>                  # block until it finishes

openart project list
openart project create --name "My Project"
openart workspace list                      # the active one is marked *
openart workspace select <workspace-id>
openart workspace select                    # back to your personal workspace

openart upload add ./reference.png          # prints a URL to use in a generation
openart upload list --type image

Scripting and agents

Three flags make the CLI safe to drive from a program:

• --json — every command emits machine-readable JSON on stdout. Pagination cursors and progress go to stderr, so a pipe stays clean.
• --dry-run — every command that writes or spends credits prints the exact request it would send, then exits 0.
• --async — generate submits and returns immediately, printing the generation's id instead of waiting for the result. Pick it up whenever you like with creation wait / creation get, so a slow video does not pin a process open. Not combinable with -o/--output, which needs a finished generation to download.

openart generate image "test" --model nano-banana-2 --dry-run

# submit now, collect later
id=$(openart generate video "a paper boat" --model pixverseV6 --async)
openart creation wait "$id"

openart model cost --json | jq -r '.items[] | "\(.model)\t\(.totalCredits)"'
openart creation list --json --limit 5 | jq -r '.data[].url'

Commands exit non-zero on failure and print the reason to stderr.

Run openart --help, or openart <command> --help, for the full surface.

Uninstall

rm "$(command -v openart)"
rm -rf ~/.openart          # also removes the stored credential

On Windows, delete %LOCALAPPDATA%\Programs\openart and remove it from your user PATH, then delete %USERPROFILE%\.openart.

Support

Bugs and feature requests: open an issue.

Include the output of openart version — it reports the exact build, platform, and commit.

License

The openart binaries are distributed under the MIT License.