The current system for crowd/community is not drawing leverage from the latest translation tools. I think we need an approach which is able to do that. We probably have several community members who are willing to sacrifice part of their LLM quotas to work on translations, but the interface needs to improve. For example:

1. Query strings (by language, status e.g.)
2. Download json export (including instructions for user/LLM, context, and translations/suggestions already made)
3. User passes through LLM of choice
4. User reviews
5. Upload as suggestion (or definite, if moderator)

Optional: Feature on Hub to mark source strings as ”Needs review, and a reason+comment”

We hope to automate the steps 1-3 and 5 on our side without users needing to do anything, we are mostly there now. We still need to rely on humans to review and approve these machine translations though.