VirtuALL
The live multilingual room, captions and record.
The problem
Live deliberation across languages needs understandable speech, chat and a traceable record.
Place in the process
Deliberate · Record
Current capabilities
Runs at live.globall.org in the 24 official languages of the European Union, all directions. Everyone may speak at once, each in their own language; captions and chat are one record. Sign-in is the GlobALL account, and roles come from membership rather than from what a browser claims. The record exports as a readable transcript, as data, and as a CanonicALL process record another tool can read.
Current limits
Nobody outside the team has used it: no language pair has been accepted by a native reviewer, and concurrent speech has not been validated by participants. Irish can be read and written but not spoken, because no speech recogniser we may ship supports it. An interpreter can hold a language channel but has no separate audio feed. Speech recognition runs on a non-EU hosted model.
Next development gate
A facilitated session with six people in three languages, following docs/six-person-test.md: does the record attribute every sentence to the person who said it, and would they call it a fair account?
Evidence record
Source: modules/virtuall/docs/status.md. 346 service checks, 125 interface checks and 31 end-to-end checks pass. Demonstrated behaviour is not the same as user validation, and nothing here has been validated by users.