Connecting API keys
Connect fal, OpenAI or OpenRouter for optional cloud AI.
Cloud features use your own provider account. Keep keys in Settings, rather than prompts, recipes or support screenshots.
fal and OpenAI images
- Open Settings → Integrations.
- Paste the image provider’s key and save.
- Choose that provider in Create or an AI Canvas step, then select a compatible model.
- Run the generation explicitly. Saving a key does not start a request.
Desktop keys use encrypted storage on this Mac. The Canvas receives connected states, not the saved keys. Remove a key in the same settings section.
fal image modes
Nano Banana 2.1 and Nano Banana 2 support generation/editing, FLUX Schnell text-to-image, and FLUX Kontext Pro reference editing. Controls adapt to supported parameters. Connecting fal does not expose its full catalogue.
fal video
The same fal key enables Seedance 2.5 in Create → Video. Generation uses your fal account credits; the prompt and optional start frame are sent to fal when you press Generate video. Video controls and progress.
OpenRouter chat and vision
OpenRouter is a separate connection for Ask and other chat/vision requests. In Settings → Integrations, save your OpenRouter key and choose an OpenRouter model ID in the Model field. When no model is saved, the initial choice is GPT-5.4 mini (openai/gpt-5.4-mini). Existing saved model choices are preserved.
Under Default for AI, choose On-device or Cloud. Cloud requests use your OpenRouter account credits and send the prompt and selected chat context or image to OpenRouter. This setting does not change the image and video models offered by Create or Canvas.
Model download authentication
A Hugging Face token authenticates model downloads; it is not an inference provider key.
Errors
Check authentication, balance, model access, rate limits and internet access. If a paid submission was interrupted, check the provider dashboard before repeating it.