OpenAI gpt-transcribe is stream capable

[rodriciru]

Hi!
OpenAI’s gpt-transcribe model supports streaming file transcription and emits transcript.text.delta events as it processes the audio. However, generateTranscription({ stream: true }) currently sends the completed transcript as a single CUSTOM generation:result event.

For long recordings, such as a 60-minute audio file, the client receives no transcript text until processing is complete.
It looks like this may be because TranscriptionAdapter.transcribe() only returns a Promise.

Could TanStack AI add a streaming transcription adapter API—for example, openaiStreamTranscription—that forwards transcript deltas as they arrive?