Skip to content

Settings: Image Generation

The Image Generation panel at Settings -> Image Generation controls Wayland’s built-in image generation tool. The tool runs as a bundled MCP server. Enable the server, pick a model arm, and Wayland syncs the configuration to your agents automatically.

The Settings Image Generation panel with the enable toggle and the default image model selector

ControlDescription
Enable toggleTurns the built-in image-generation MCP server on or off. Enabling syncs it to your agents. Disabling removes it from all agents. The toggle is unavailable until an image-capable model is selected.
Agent status indicatorWhen the server is on, shows which agents have the image tool installed.

The model picker lists available image models grouped by connected provider. Only providers that expose image-capable models appear here.

Selecting a model writes the provider connection details (provider, base URL, API key, and model id) into the server environment. The server uses those details for every generation request until you change the selection.

When Flux Router is connected, the picker offers the full set of Flux image arms. The default selection is Flux Image (flux-image), which lets Flux Router pick the best-performing arm per request. Power users can pin a specific arm.

Arms available when Flux Router is connected, in picker order:

Arm idLabelNotes
flux-imageFlux ImageDefault. Flux picks the best arm per request.
gpt-image-highGPT Image (High)Requires a verified OpenAI org behind Flux.
nano-banana-pro-4kNano Banana Pro 4KGoogle Gemini-backed. SynthID watermark applies.
nano-banana-pro-2kNano Banana Pro 2KGoogle Gemini-backed. SynthID watermark applies.
gpt-image-high-xlGPT Image (High XL)Requires a verified OpenAI org behind Flux.
nano-bananaNano BananaGoogle Gemini-backed. SynthID watermark applies.
gpt-image-medGPT Image (Medium)OpenAI medium quality.
flux-image-together-fluxTogether FLUX (Fastest)Fastest, lower fidelity.

When a non-Flux image-capable provider is connected (for example, a native OpenAI account or an OpenRouter connection), the picker lists the image model ids that provider exposes. A curated floor of known good ids is shown by default; any additional ids the provider reports are added on top.

When no image-capable models are available, the panel shows a link to the image generation setup guide. Connect an image-capable provider first, then return to this panel.