Changelog
What has actually changed, newest first. PicsTag is maintained by a small studio, so this moves in bursts rather than on a release train.
Documentation, diagrams and bylines
- The app page now carries a written reference for the CSV input format, the export schema and the common failure modes — previously this only existed in scattered FAQ answers.
- Added diagrams to the alt text and metadata guides: a side-by-side of four alt texts for the same photo, and a map of where EXIF, IPTC and XMP actually sit inside a file.
- Guides are now grouped by topic, carry a reading time, and link to their siblings.
- Every guide has a named author and a byline.
First public release
- The tool: drag-and-drop images or a CSV of URLs, tag and caption them with models running locally in the browser, review each tag, export to CSV or JSON.
- Optional OpenRouter provider for people who want a more capable cloud model and are happy to bring their own API key.
- WebGPU with an automatic WebAssembly fallback, and 8-bit quantised models so the first-run download is about 250 MB rather than a gigabyte.
- Fifteen guides on image metadata, alt text and keywording.
Missing something you need? Tell us what you are trying to do — feature requests that describe the problem rather than the feature tend to be the ones that get built.