Nano Banana (Gemini's image generation/editing model) does not honor a caller's desired output aspect ratio on its own. When it generates or edits an image, the result inherits the aspect ratio of the uploaded input image — a square input stays square, a landscape input stays landscape — regardless of what aspect-ratio option a surrounding app's UI offers the user.
This surfaced building a Nano Banana-powered product-photography app ('Product Banana') in Google AI Studio's code-assistant builder: selecting an aspect ratio in the app UI had no effect until the app was explicitly instructed to reformat the uploaded image to match the selected ratio before sending it to Nano Banana.
Apply: When building an app or wrapper around Nano Banana (or similar image models with the same input-preserving behavior) that exposes an output-aspect-ratio control, don't rely on passing the aspect ratio as a prompt parameter alone — add an explicit pre-processing step that crops/reformats the uploaded image to the target aspect ratio before the model call. See Node-Based Prompt→LLM→Image→Video Pipeline Chaining for the general pattern of inserting deterministic pre/post-processing steps around a model call in a pipeline.
Из тем: Workflow Tooling: Node-Based Pipelines, Claude, and Multi-Model Chains