Lexica

Indexes millions of Stable Diffusion-generated images with their prompts and metadata, enabling full-text and semantic search across the corpus. Uses embedding-based retrieval to match natural language queries against stored prompt embeddings and image features, returning ranked results with generation parameters (model version, sampler, seed, CFG scale). The search engine crawls public Stable Diffusion outputs and maintains a searchable database of prompt-to-image mappings.

Parses and displays the complete generation parameters for each indexed image, including model checkpoint, sampler algorithm, guidance scale, seed, steps, and any LoRA or embedding modifications. Extracts this metadata from image EXIF data, generation logs, or associated metadata files, then presents it in a structured format that users can copy directly into their own generation tools. This enables direct reproduction or modification of successful generations.

Maintains a bidirectional index linking natural language prompts to their generated images, allowing users to search by prompt text and discover all images created from similar or identical prompts. Uses full-text search on prompt strings combined with semantic similarity matching to surface variations and related prompts. Aggregates multiple generations from the same prompt to show consistency or variation in outputs.

Aggregates engagement metrics (views, likes, shares, saves) from the Lexica platform and community sources to identify trending generations and popular prompts. Ranks images by recency, popularity, and quality signals, surfacing high-engagement outputs on discovery pages and trending sections. Uses collaborative filtering or engagement-based ranking to promote community-favorite generations without explicit user ratings.

Provides faceted search filters allowing users to narrow results by specific generation parameters: model checkpoint, sampler type, guidance scale range, step count, aspect ratio, and other technical attributes. Implements multi-faceted filtering where users can combine constraints (e.g., 'DPM++ sampler AND 7.5 CFG AND 768x512 resolution') to find images matching specific technical criteria. Filters are applied server-side to the indexed corpus, returning only matching results.

Maintains creator profiles that aggregate all generations uploaded or shared by individual users, displaying their generation history, favorite prompts, preferred parameters, and community engagement metrics. Profiles show patterns in a creator's work (favorite subjects, consistent aesthetic, parameter preferences) and enable following creators to discover their new generations. Tracks creator reputation through community metrics and generation quality indicators.