The agent can apply denoise, deblur, color correction, white balance adjustment, exposure correction, face enhancement, lighting improvement, and clean-up enhancements. Multiple enhancements can be combined in a single API call. Industry presets like e-commerce and real_estate bundle the most effective combinations for those use cases.

The agent passes the image URL to Deepimage's processing endpoint with the background parameter configured. You can remove the background entirely for a transparent PNG, replace it with a solid color, or generate a new background using AI. The processed image returns within seconds for the customer to review.

Deepimage can upscale images up to 16x their original size or up to 300 megapixels. You control the output dimensions using width and height parameters in pixels or percentages. The output_format parameter lets you choose PNG, JPEG, or WebP, and the quality parameter adjusts compression from 0 to 100.

The synchronous endpoint returns results if processing completes within 25 seconds. For larger images, it returns a job hash instead. Your agent then polls the Get Job Result endpoint using that hash to retrieve the final image when ready. This prevents conversation timeouts while still delivering the enhanced result.

Tars sends image URLs to Deepimage for processing and receives result URLs back. The processed images are stored on Deepimage's servers temporarily. You can use the Delete Job endpoint to remove processed results when no longer needed, or configure a target storage destination for output images.

Yes. Deepimage offers preset configurations including real_estate and ecommerce that bundle optimized enhancement combinations. The e-commerce preset focuses on product clarity, background cleanliness, and color accuracy. The real estate preset emphasizes lighting correction and spatial enhancement. Pass the preset parameter to apply them.

Basic resizing stretches pixels and creates blur. Deepimage's generative upscale uses AI to create new image details that match the original content. For example, it can add realistic texture to foliage, fabric detail, or skin in portraits. This produces sharp, detailed enlargements rather than fuzzy scaled-up versions.

The agent can submit multiple images sequentially using the async processing endpoint. Each submission returns a job hash for tracking. Your agent can queue several images and retrieve results as they complete, enabling batch workflows. Check your account credits beforehand using the Account Info endpoint to avoid interruptions.