Libraire

Searches a curated library of millions of AI-generated images using natural language queries and visual similarity matching. The system likely indexes images with embeddings (CLIP or similar vision-language models) to enable semantic search beyond keyword matching, allowing users to find visually similar images or images matching descriptive text prompts without exact tag matches.

Enables downloading multiple images from search results or collections in batch operations, likely with options for format conversion, resolution selection, and metadata export. The system probably queues downloads server-side and provides a manifest or archive (ZIP) containing images with standardized naming and optional JSON metadata (prompt, generation model, creation date).

Allows users to create, organize, and share custom collections of images from the library through a tagging and folder-like organizational system. Collections likely support collaborative access control, allowing teams to curate shared mood boards or asset libraries with role-based permissions (view-only, edit, admin) and version history for collection changes.

Accepts an uploaded image or image URL and returns visually similar images from the library using CLIP-style vision embeddings or perceptual hashing. The system compares the input image's embedding against the indexed library and ranks results by cosine similarity, enabling users to find images with matching composition, color palette, or visual style without needing text descriptions.

Stores and exposes generation metadata for each image in the library, including the original prompt used to generate it, the AI model/version that created it, generation parameters (seed, guidance scale, steps), and creation timestamp. This metadata is likely queryable and exportable, allowing users to understand how images were created and potentially use prompts as inspiration for their own generation workflows.

Provides multi-dimensional filtering across image attributes such as generation model, creation date range, image dimensions, color palette, aesthetic style, and content tags. Filters are likely applied server-side with faceted search UI showing available filter options and result counts, enabling rapid refinement of large result sets without re-querying the full library.

Exposes REST or GraphQL API endpoints for querying the image library, retrieving search results, accessing metadata, and managing collections programmatically. The API likely supports pagination, filtering, sorting, and bulk operations, enabling developers to integrate Libraire into applications, build custom search interfaces, or automate asset pipelines without relying on the web UI.

Provides explicit, standardized licensing information for all images in the library, likely under a single unified license (e.g., CC0, custom commercial license) that applies to all AI-generated content. The system eliminates per-image licensing complexity by guaranteeing that all images have identical usage rights, removing the need for license verification or attribution tracking that plagues traditional stock photo services.