Google Cloud Object Enrichments is a Cloud Storage preview that asynchronously processes eligible images and PDFs with selected models, then attaches the results to each object as structured context. The metadata can support later search and analysis; Google makes the feature available through the Cloud Storage console and API. Google Cloud’s overview
Check both the model task and the bucket setup: a useful task still depends on an eligible object in a supported bucket.
What the five model tasks return
Object Enrichments offers five documented tasks. Their outputs differ, so the right choice depends on whether you need image labels, location details for detected objects, face-related data, content ratings, or PDF categories.
| Model | Documented task or output |
| Object detection | Identifies objects and returns their positions, including rectangular boundaries. |
| Image labeling | Returns labels that describe image content. |
| Face detection | Can return bounding polygons, facial landmarks, orientation angles, and expression likelihoods. |
| Explicit content detection | Returns likelihood ratings for adult or violent content. |
| PDF classification | Assigns categories to PDFs. |
Google documents these tasks and outputs in its Cloud Storage overview. Processing is asynchronous: results are attached to the stored object as object context under contexts.custom.
Where the context goes
Object context keeps the generated metadata associated with its Cloud Storage object. Google says access can be governed with IAM controls, and its documentation describes using Storage Insights datasets to export the context for analysis in BigQuery. That gives teams a route from stored files to downstream queries without making the object-context metadata itself a separate model task. Google Cloud documentation Storage Insights datasets documentation Google Cloud announcement
Google sets a maximum JSON payload of 128 KiB for an object context. Google Cloud announcement
Which files and buckets qualify
An image or PDF must be no larger than 20 MiB and have a valid, supported Content-Type. Google lists BMP, GIF, ICO, JPEG, PNG, TIFF, and WEBP as supported image formats; PDFs must use application/pdf. The processing locations are us, eu, and asia. Configuring the same bucket and model in multiple locations repeats processing and incurs a charge for each run. Google Cloud’s overview
The preview does not support Rapid buckets, buckets with hierarchical namespace enabled, or objects encrypted with customer-supplied encryption keys. Configurations use exact bucket names rather than wildcards or regular expressions. Bucket relocation is incompatible with Object Enrichments, and a newly created or recreated bucket is not automatically added to an existing configuration. Those details matter when a team changes its bucket inventory or moves data after setting up enrichment. Google Cloud’s overview
Gemini is outside the preview
Native Gemini enrichment is not part of the Object Enrichments preview. Google describes it as forthcoming functionality and says interested customers can request access. The five tasks currently described for the preview are the ones in the table above. Google Cloud announcement