Thesis
The integration makes Microsoft AI audio models accessible through Vercel AI Gateway, adding a managed route for speech generation and streaming transcription.
Observed facts
Vercel lists MAI-Voice-2.1 and MAI-Voice-2.1-Flash for speech generation, and MAI-Transcribe-2 Streaming for partial transcript updates as audio arrives.
1Its changelog includes AI SDK examples using
1generateSpeechandstreamTranscribe, and says Gateway charges listed rates without platform fee or inference markup.
Mechanism
Applications can call the named models through AI Gateway and use the SDK’s speech-generation or streaming-transcription functions. Partial transcript updates let an interface display interim text while audio is still arriving; the source cautions that partials can change and should be replaced as new ones arrive.
Why now
Vercel has announced the models as now available, so developers evaluating voice features can consider this route in current implementation work.
Implications
Teams already using AI Gateway can evaluate speech output and live transcription without integrating a separate provider path described in this announcement.
The stated no-markup billing may simplify cost accounting, but does not establish that the models are cheaper overall.
Evidence boundary
The source is Vercel’s changelog and documents availability, examples, and billing terms; its security and zero-data-retention language is partner positioning, not independent validation, and it provides no benchmark or deployment outcome.
Watch conditions
Check for independent latency, transcription accuracy, language coverage, and cost comparisons under representative workloads, and verify retention and security terms against applicable documentation before choosing a production workflow.