Speech input and translated text are different
An input language is the language spoken into the microphone. A translation pair describes the text produced in another language. Recognizing speech in a language does not by itself establish a translation pair.
The first planned pairs are French → English and English → French. Live Caption Studio is built for speakers who switch between French and English without choosing a language. We are not publishing an accuracy claim for those switches yet.
The language roadmap
French, English and FR↔EN translation are the first planned beta scope. Spanish is next; Italian, then Portuguese and Haitian Creole follow in the plan. There are no promised dates for these additions.
The later languages are roadmap items. We do not claim additional translation pairs at version 1.0, or additional input-only beta languages before their evaluation. Model makers’ coverage lists are not our list of supported languages.
How a language earns support
A language changes status only after our evaluation passes for it. We measure speech recognition on the relevant recordings and assess translation for understandable, faithful meaning; bilingual switching also needs its own evaluation.
When results are ready, we will identify the test material, hardware, date and method. Until then there is no measured accuracy percentage, delay figure or claim that all accents and rooms perform alike.
Text for the screen, stream and phone
The planned translation output is written text alongside captions. It is intended for audience screens, stream overlays and the phone viewer. Translated audio is a later roadmap item, not a version 1.0 feature.
A bilingual conference or church service still needs a clean speech feed from the console. A second language does not remove the audio requirement: speech over a music bed carries no accuracy promise.
Tell us what your room speaks
The beta form asks which languages are spoken in your room and lets you name another language. Those answers also act as votes for the next languages.
Use the existing notes box (200 characters) to describe the text language you need and whether speakers switch during a sentence. This helps describe the use case; it does not confirm a support date.