Generate or edit an image from the command line with no API key, free tier of kavel.ai
311
docker run --rm san8888/kavel "an isometric coffee shop, pastel palette"
# https://cdn.kavel.ai/uploads/kie/image/....webp
A 6 MB scratch image with one static binary. No key, no account, no environment variables: it talks to
the free tier of Kavel, an AI image and video studio, using a client id
it makes up per run.
# generate, with an aspect ratio
docker run --rm san8888/kavel -ratio 16:9 "matte black ceramic mug on pale oak, soft window light"
# edit an image you already have (public http(s) url)
docker run --rm san8888/kavel -edit https://example.com/portrait.jpg "give him a short textured crop, keep the face"
# use the url in a script
URL=$(docker run --rm san8888/kavel "flat illustration of a rocket, purple and orange" | tail -1)
Past the free tier, pass an API key from kavel.ai/settings/apikeys and pick any model your plan includes — calls then run on your account, with no per-IP ceiling and no watermark on a paid plan:
docker run --rm -e KAVEL_API_KEY san8888/kavel -model gpt-image-2 "a product shot of a glass perfume bottle"
Runs on linux/amd64 and linux/arm64. Handy in CI jobs, cron containers and anywhere installing a
language runtime just to make one picture is overkill.
Signing in on kavel.ai removes the watermark and opens the larger models — Nano Banana 2, GPT Image 2, Seedream 5.0 Pro — and the AI video generator. Pricing.
Say what changes and what stays. The edit engine, Nano Banana 2 Lite, keeps what it can still see and invents what it cannot. Tuned prompts to copy from: the AI hairstyle changer and the AI outfit generator.
Built from kavel-go. SDKs for Python, Rust, TypeScript, PHP, Ruby and Dart, plus a GitHub Action: kavel.readthedocs.io.
MIT
Content type
Image
Digest
sha256:9a0a6eef4…
Size
2.9 MB
Last updated
23 days ago
docker pull san8888/kavel