Made together
About Community Dictionaries
Build language resources from the pictures, voices and experiences that matter to your group.
Start with everyday life
Take a photograph or upload one, add words or a sentence, and record someone saying them. People can contribute different parts: one person supplies a picture, another the language and voice. Pictures, sentences and word pages link together so you can explore them in either direction.
No programming or software installation is needed. Use a phone for photographs, recordings and browsing, or a laptop when you prefer a larger screen for editing. Creating a dictionary, manual contributions and browsing have no platform charge under the current hosted arrangement.
Your language, your choices
Human pictures, text and recordings work without AI support for the language. Where suitable AI support is available, you can optionally ask for picture descriptions, translations, speech or generated illustrations. An owner can add missing sentences to a collection of pictures, then build vocabulary from the reviewed sentences.
AI suggestions can be wrong and should be checked by someone who knows the language. Optional AI processing can spend credits and send selected material to a provider; the relevant tools explain their processing and costs. It is not required for manual dictionary work.
Members, and visitors to dictionaries with public browsing enabled, can practise with flashcards, picture crosswords and word-search puzzles using the dictionary's existing material. These exercises make no paid AI calls.
Decide what to share
Dictionaries start private and unlisted. Owners can make a dictionary discoverable, showing its name, languages and accepted entry counts so people can request membership. This does not expose its contents.
Public browsing is a separate owner choice, made with contributor permission. It allows anyone with the link to browse accepted learning material. A public dictionary only appears in the discovery list if it is also discoverable.
Contributors can withdraw all their contributions from a dictionary into their private collection and later restore them. Withdrawal blocks subsequent access to the hosted material, but cannot recall copies people already received. Hidden dictionaries are not advertised. Your personal contributor listing is optional too.
A project that learns from its users
Community Dictionaries is part of the open-source C-LARA project. AI has written the application code and technical documentation through an iterative collaboration with people who propose requirements, install releases, test them and give feedback. The aim is to make useful changes quickly while keeping the app understandable.
The project is still developing. We are investigating how well this approach serves different language communities, and how much technical work AI can reliably take on over time. Independent community trials and long-term results are still needed.
Publications
Community Dictionaries: Everyday Pictures, Rapid Human–AI Development
ChatGPT C-LARA-Instance and Manny Rayner (2026).
Our first report describes the working app, its development and an invitation to help shape the next version.
Read on AiraXiv · Read on ResearchGate
Related reading: Vibe Research, by Manny Rayner and ChatGPT C-LARA-Instance, discusses the broader human–AI research collaboration.
Contributors
People who have chosen to be listed. Names and introductions are supplied by the account holders. This is a community listing, not the author list of a publication.
Manny Rayner — I am at the moment the main technical person in this project. But to be honest, GPT-6 Astra does nearly all the work.
Choose List me as a contributor in Profile to appear here.
Questions and suggestions
Tell us what would make the app more useful for your group, or ask about trying it together. Contact the project.
Feedback currently goes to the project contact by email. There is no public forum or automatic AI moderation of these messages.