EchoFlow Nano Banana Pro Image Generation and Editing

Problem

When a workflow needs to turn natural-language prompts, existing images, or multiple source images into image results, hand-written HTTP calls can become repetitive and error-prone. This skill wraps the EchoFlow API image capabilities into a repeatable CLI script, reducing ad hoc debugging around authentication, parameters, output filenames, and resolution behavior.

How It Works

The skill runs a bundled Python script with uv, which manages dependencies such as httpx and pillow. It prefers the ECHOFLOW_API_KEY environment variable and calls the EchoFlow image endpoint. Core capabilities include generating an image from --prompt, editing a single image with --input-image, and composing up to 14 input images into one result. --resolution supports 1K, 2K, and 4K; during editing, resolution can be auto-selected from the largest input image size. The default model favors faster and more stable preview behavior, while a higher-quality preview model is also available.

Boundaries and Notes

It depends on a valid third-party API key, network access, and EchoFlow service availability, making it suitable for local debugging, automation, or agent toolchains rather than offline use. Keep secrets in trusted environment variables instead of CLI arguments. If --api-base is changed, the key is sent to the custom host, so only point it to a trusted endpoint. Multi-image composition is limited to 14 inputs, and higher-quality models may hit 429 rate limits.