A11yGen: open-source generative AI tool automates context-aware alt text for large image catalogs

AI · 4 min read

A11yGen, released August 5, 2026, combines a lightweight on-premise transformer with rule-based heuristics to generate alt text for large image catalogs. Unlike generic captioners, A11yGen allows teams to provide contextual metadata (page intent, audience, important details) so generated descriptions align with the content's purpose and privacy constraints.

Key features include adjustable verbosity (brief, explanatory, and detailed), a 'safety-first' filter to avoid sensitive identifications, and hooks for human review workflows. The project emphasizes enterprise adoption: it supports batch processing, integrates with DAMs, and can run entirely offline to meet data residency needs.

Early adopters at e-commerce and publishing companies report a significant reduction in alt-text backlog and consistent improvements in screen reader user satisfaction when combined with a lightweight editorial QA process. Maintainers invite contributions and are working on a plugin ecosystem to adapt generation rules to sector-specific needs like medical imaging and cartography.