v1.7.0 — Media nodes, GPT Image, and canvas node search

Node Banana 1.7.0 adds new media nodes, GPT Image and video-to-video support, a searchable canvas menu, and a broad reliability pass.

New nodes

• Image Resize: resize or scale an image to exact dimensions, with a maximum-size guard.
• GIF Encoder: combine frames or images into an animated GIF.
• Remove Background: one-click background removal that runs in the browser.

Models

• GPT Image (OpenAI): image generation and editing with gpt-image-1 and gpt-image-2 through the OpenAI Images API. Set an OpenAI API key to use them.
• Replicate model search now returns the full catalogue instead of a subset, so every Replicate model is discoverable.
• Generate nodes accept a video or audio input, so video-to-video models (such as upscalers) and image- or audio-only generation run without a text prompt.

Canvas

• Double-click or right-click an empty area of the canvas to open a searchable list of every node type, with keyboard navigation.

Gallery and output

• Output and video galleries use precomputed thumbnails for faster browsing.
• Multiple and looped video outputs are collected into the gallery.
• Removing a gallery item removes the one you clicked, even when duplicates are present.

Reliability and performance

• A single node throwing no longer takes down the app; failures are contained with an inline recovery option.
• Fixed a data-loss race when editing during an auto-save.
• More reliable resume after a pause edge.
• Stop now cancels in-progress video stitch, trim, and ease-curve encodes immediately.
• Oversized generated media is handled gracefully.
• Bounded caches, more resilient provider polling and downloads, and dependency-aware concurrent execution.

Full changelog: v1.6.0...master