apiwatch

All changes / OpenAI

Breaking rest

OpenAI shuts down /v1/audio/translations on February 26, 2027; the replacement is Realtime translation at /v1/realtime/translations, which needs a separate integration

Takes effect · Audio translations endpoint (/v1/audio/translations)

Reviewed against the vendor source on Nov 19, 2026 (AI-assisted review). Key facts are re-checked against the source daily.

What changes

Notice date August 26, 2026. OpenAI's deprecations page lists /v1/audio/translations as Deprecated with a February 26, 2027 shutdown date and /v1/realtime/translations as the recommended replacement. The API reference says the endpoint supports only whisper-1, that the endpoint and model are scheduled to shut down on February 26, 2027, and that /v1/realtime/translations requires a separate integration. The Speech to text guide says the endpoint translates a completed recording into English text only, and that changing only the model ID in a /v1/audio/translations request doesn't migrate it: Realtime translation with gpt-realtime-translate needs a translation session, streamed audio input, and translation events. whisper-1 itself is covered by the transcription models record with the same date.

Breaking: Calls fail after this date.

What to do

Rebuild audio translation on Realtime translation (gpt-realtime-translate at /v1/realtime/translations) using OpenAI's Realtime translation guide. It streams audio into a translation session and returns translated audio and transcript deltas, so a model swap alone doesn't migrate a /v1/audio/translations call; set the session's target language and re-test on your own recordings before February 26, 2027.

Strings to search for

/v1/audio/translationsaudio.translations.create

Not affected (matches inside these are ignored): api.groq.com/openai/v1/audio/translations

Counts only in files that mention openai.

Source

https://developers.openai.com/api/docs/deprecations