Photos, scans and screenshots often hold information that never makes it into a file name. AI Image Captioning closes that gap. Egnyte automatically generates a detailed description of every newly uploaded image and saves it on the file, so image content can be found through search and used by the AI Assistant and Agents in the same way as text in a document.
Image captioning is part of Egnyte's Context Layer, which automatically prepares content for AI by classifying, enriching and connecting files as they arrive. Captions require no setup and no manual tagging.
- The feature is available to all the customers with the AI add-on enabled. Please contact the Egnyte account manager or Egnyte Sales team for more information.
- The supported file types are PNG, JPG / JPEG, TIF / TIFF, BMP, JPX, GIF, WEBP, and HEIC with the maximum file size as 64 MB per image.
Use Cases
- Construction and Engineering: Find job site photos by condition, equipment or progress stage, and gather visual evidence for RFIs, inspections and daily reports
- Marketing and Creative Teams: Locate brand assets, product shots and event photos by what they show
- Facilities and Real Estate: Search for property and maintenance photos showing a specific room, fixture or issue
- Insurance and Claims: Locate damage photos by type of damage or affected item
- Scanned Notes and Signage: Find whiteboard photos, labels and signs by the text written on them
How Captions are Generated
When an image is uploaded to Egnyte through any channel, such as the Web UI, the Desktop App or an integration, Egnyte sends it to an AI vision model for processing. The model examines the image and writes a description covering:
- Objects and Equipment: for example, a yellow bulldozer, a fire extinguisher or a laptop
- Scene and Setting: such as, a dirt construction site, an office lobby or a warehouse aisle
- Materials and Conditions: for example, exposed rebar, standing water, mud on equipment or a cracked surface
- Visible Text: words on signs, labels, banners, whiteboards and equipment markings
- Notable Absences: where relevant, the caption notes what is not present, for example, no workers or safety equipment are visible
The caption is saved on the file as an Egnyte Document Tag named Caption and is added to the search index. Users can open the image and open the Tags panel to view the caption.
Captions are generated shortly after upload. During periods of high upload volume, some images can take longer to be captioned, but every supported image is processed.
Captions describe people in general terms, for example, a person in a high-visibility vest. They do not identify individuals by their face, and Egnyte does not perform facial recognition on images.
Generating a Caption for an Existing Image
Images that were already in Egnyte before the feature was enabled are not captioned automatically. Users can generate a caption for an existing image on demand:
-
Open the image in the file preview and open the Tags panel. Select Auto tag. Users can also select multiple files and add captions using Egnyte’s multi-document tagging.
The multi-document tagging feature is available for customers on Business, Enterprise Lite, Enterprise, Essentials, Elite and Ultimate plans. To request access, contact the Egnyte Products team.
- Select Egnyte Document Tags and the caption will automatically be generated. Click on Save.
The caption is added to the Tags panel and to the search index within a few moments.
How Captions are Used
Search
Captions make image content searchable alongside documents. Users can find an image by describing what it shows instead of remembering its file name or folder. For example, a search for bulldozer on a dirt site or whiteboard with a sprint plan returns matching images even when the files are named IMG_4821.jpg.
In the example below, a search for photos with steel rebar bundles in a site photos folder returns the matching image. The AI Overview describes what the photo shows, using the image caption.
Egnyte's AI Assistant
The AI Assistant uses captions to find relevant images and to answer questions about them. Users can ask questions such as:
- Find photos in the Riverside project folder that show water damage.
- Which site photos from last week show the crane?
- List the event photos that show the company banner.
The AI Assistant can also read the image itself when a file is selected, so captions and image content work together. Users can view the source image(s) in the answers.
In the example below, a user asks the AI Assistant which site photos show steel rebar or heavy equipment. The AI Assistant lists the matching photos with a short description of each, and notes the photo it excluded.
Agents
Agents use captions to search and reason over images within the content they have access to. This lets an agent include photos in its response such as checking site photos against a punch list, gathering inspection photos for a report, or finding product images that match a description.
In the example below, a site photo inspection agent, with a site photos folder as its source, is asked to check the photos for safety issues. The agent reviews every image and returns a table with what each photo shows and any visible safety or housekeeping concerns, then recommends which issues to check on site first.
Limitations
- Captions are generated automatically only for images uploaded after the feature is enabled on the domain. Images that were already in Egnyte can be captioned with Auto tag, as described above. Bulk Tagging can only be done using Multi-document tagging available on eligible plans.
- Images larger than 64 MB are not captioned.
- Captions are generated by AI and can occasionally miss details or describe an image imperfectly. Users should verify important details by opening the image.
- Captions do not identify people by their face.
- Image captioning in the Egnyte Mobile Apps is a separate, on-device feature. Refer AI Image Labelling on Mobile.