describe-image

To generate Russian text descriptions from screenshots and charts, run the describe-image script on a single file or folder. It uses a Vision AI model to produce readable text representations of visual content.

Yes, you can process a folder of UI screenshots in a batch. The script processes folders in pages and writes an LLM-friendly manifest file for each page, allowing you to view and iterate on the per-image descriptions later.

Vision AI image description handles batch processing by paginating through folders and generating a markdown manifest for each page. This manifest provides a structured reference of all extracted visible text and descriptions.

Yes, you need a VISION_API_KEY for chart analysis with Russian output. The script requires this authentication to send base64-encoded images to an OpenAI-compatible chat-completions API payload.

Vision AI screenshot analysis supports common raster image types. The script encodes images as base64 data before sending them to the OpenAI-compatible API for processing.

To improve accuracy when generating image descriptions from diagrams, you can use optional contextual grounding. This feature provides additional context to the Vision AI model, enhancing the precision of the Russian text output.