FAQ¶

Can I use my ChatGPT subscription instead of an API key?¶

Yes. Sign in with codex login, then pass --model codex:image. Codex picks the image model and size, so use an API model when you need those controls. See Codex subscription.

Why does genimg ask me to choose a model?¶

Save a default once with genimg models set-default gdm:nb2, or pass --model each time. genimg has no built-in default model.

Where are my images saved?¶

In ~/.genimg/generations/ unless you pass --output. genimg history lists them and genimg history view browses them. Each generation also gets a JSON record (prompt, model, settings, estimated cost) in ~/.genimg/metadata/.

A model is listed in genimg models but fails to generate. Why?¶

A provider can list a model that your account, endpoint or region cannot serve. Only a small generation with that exact model proves it works.

Where did Imagen go?¶

Google retired the Imagen 4 API on 17 August 2026. Use gdm:nb2 instead. If Imagen was your saved default, run genimg models set-default gdm:nb2.

Can I use GPT Image 2.5?¶

Yes, as oai:gi2.5 (Sunburst) or oai:gi2.5-flare, with an OpenAI or Azure OpenAI key. Whether your account or deployment serves them varies; genimg models shows what it lists.

Draw Studio says Failed to fetch¶

Run genimg draw again and open the URL it prints. The tab has probably outlived its local server.

Why do Draw Studio's controls change with the model?¶

Each model accepts different sizes, aspect ratios and quality levels. Draw Studio only shows the controls the selected model accepts.

Can I use Draw Studio from an iPad?¶

Yes. Sidecar puts the Mac's screen on the iPad, or genimg draw --host serves the Studio on the Mac's Tailscale or Wi-Fi address for Safari. See Draw from an iPad.

Does Draw Studio work offline? Can I host it?¶

No to both. It runs on your machine but sends prompts and images to the provider. --host shares your running Studio with your own devices for a session; there is no hosted, multi-user version.