REST API — Audio inference
Voice (/developers/rest-api-reference/inference/voice)
Updated the language field on transcript responses: documented as a BCP-47 code (e.g. en, es-mx) instead of an empty value with language detection not yet enabled. The same wording applies to per-channel language. Example response now shows "language": "en".
Speech to text (/developers/rest-api-reference/inference/speech-to-text)
Same language field updates as Voice: top-level and per-channel values described as BCP-47 codes, with the sample response using "language": "en".
Model capabilities — Audio
Speech to text (/developers/model-capabilities/audio/speech-to-text)
Response docs and example now present language as a BCP-47 code (e.g. en, es-mx) rather than a full language name such as "English".
Build
Settings reference (/build/settings/reference)
Clarified keep_text_selection: the word_select option description now includes triple-click paragraph select alongside double-click word select.