Runware
in your terminal.
Drop in your API key and run any model with one command.
A single static binary for image, video, audio, text, and 3D, plus model search, presets, uploads, and account management.
$
Key in. First image out.
One line for your platform.
Store your key once.
Generate from any model.
$
Showing install, auth, and run flow for macOS
One command, every modality.
Give runware run a model's AIR identifier and key=value parameters. It reads the model's schema to detect the task type, then downloads the result.
$ runware run runware:400@1 positivePrompt="A woman wearing flowers on top of her head." width=1024 height=1024saved path=outputs/1253900f….jpg
![Woman wearing flowers on her head, generated with FLUX.2 [dev] (runware:400@1)](/_next/image?url=https%3A%2F%2Fassets.runware.ai%2Fexamples%2Fbfl-flux-2-dev%2Ff801dc92-7c53-4494-9279-56e61a3b70c1.jpg&w=3840&q=75)
1253900f….jpg · 1024×1536
$ runware run xai:[email protected] positivePrompt="Slow forward camera glide beside the paper train as it crosses the trestle, subtle folds fluttering along the carriages, warm dawn rays shifting across the ravine, koi-shaped clouds drifting in graceful spirals" width=1280 height=720 duration=12 inputs.frameImages.0="https://assets.runware.ai/assets/inputs/f0eb5a7f-4fe8-4cd3-9aef-322b55e44904.jpg"saved path=outputs/f9a79214….mp4
$ runware run minimax:[email protected] positivePrompt="Instrumental only, no vocals, no lyrics. A crystalline electronic ambient track with slow pulse, icy synth textures" settings.instrumental=truesaved path=outputs/35af8646….mp3
$ runware run google:[email protected] messages.0.role=user messages.0.content="Analyze the provided image like a meticulous visual investigator. Return four sections with clear headings: Scene Summary, Concrete Evidence Observed, Most Likely Recent Sequence of Events, and Uncertainties and Alternate Interpretations." inputs.images.0="https://assets.runware.ai/assets/inputs/c5c8c57a-3337-4cb9-803a-c14a92b0742c.jpg"
$ runware run meshy:meshy@6 positivePrompt="An ornate tabletop music box shaped like a leaping koi fish wrapped around a circular porcelain base, elegant fins forming structural supports, brass winding key, enamel inlays"saved path=outputs/bd52d3fe….glb
Showing Image example
Every new model, the day it ships.
run reads each model's schema at call time, so a model works the moment it's live on Runware. No CLI update, no waiting. The whole catalog is one command away, from major providers to thousands of community models.
400K models across every modality
- Nano Banana 2
- FLUX.2 [klein] 9B KV
- GPT Image 2
- Krea 2 Turbo
- ImagineArt 2.0
- TwinFlow Z-Image-Turbo
- Seedance 2.0
- Kling 3.0
- PixVerse V6
- HeyGen Avatar V
- HappyHorse-1.0
- Runway Aleph 2.0
- DeepSeek V4
- MiniMax M2.7
- SAM 3D Objects
- ACE-Step v1.5 Base
- Inworld TTS-1.5 Max
- Gemini 3.1 Flash TTS
- Veo 3.1
- LTX-2
- Wan2.7
- GLM-4.7
- TRELLIS.2
Built to pipe.
Every command prints a table, or JSON and YAML for machines. Logs and progress go to stderr, so stdout stays clean for piping. Set RUNWARE_API_KEY and it drops straight into CI.
Pipe the output
$ runware run runware:400@1 positivePrompt="Product hero" width=1024 height=1024 --format json --no-download | jq -r '.imageURL'
https://im.runware.ai/image/os/.../1d0c0c3a.jpg
Run it in CI
The whole API, one binary.
More than generation. Search the catalog, upload your own models and assets, save presets, and manage your account, all from the terminal.
$ runware run runware:400@1 positivePrompt="A woman wearing flowers on top of her head." width=1024 height=1024saved path=outputs/1253900f….jpg
Showing terminal output for run
Questions, answered.
From install to first generation in minutes.
Install, authenticate, run.