Translations are produced by Anthropic's Claude with automatic backup from GPT-4o (OpenAI), two of the most capable AI systems available. This page shows plainly what those systems can and cannot deliver as of July 5, 2026. Every delivery is checked by a second AI and scored for translation quality. The table below shows each language's coverage: text, spoken audio, and model quality.
| Language | Text | Spoken aloud | Model quality |
|---|
Text: available in every listed language. After delivery, each translation is scored by a second AI: Claude (Anthropic) translates, and an OpenAI model reviews it, comparing the translation to the original and scoring how much of the meaning survived, from 0 to 1. Low scores are flagged on the delivery record.
Spoken aloud: device speech voices do not exist for every language, because phone makers such as Apple and Google have not built voices for them. Where a voice is missing, spoken audio arrives as captions with a note rather than failing silently. The voice is missing from the phone itself; there is no setting that can turn it on.
Sign languages: leading AI labs, including Anthropic, OpenAI, and Google, cannot produce reliable sign language video as of July 5, 2026. American Sign Language is delivered as written captions for now. Video needs to be developed with Deaf-led evaluation, and production SMEs.
Emerging: works today; have a human linguist SME verify important messages.