vercel/ai @ai-sdk/azure@3.0.131 on GitHub

b98e43b: Route mai-transcribe-1.5 to the Azure Speech API by default, like mai-transcribe-2. MAI-Transcribe-1.5 requests no longer send the segment timestamps default, which the model rejects.
2443982: Add MAI-Voice-2-Flash and MAI-Voice-2 speech generation through azure.speech() using Azure Speech text to speech (SSML), with voice, output format, speed, and style/styleDegree provider options. language picks a default voice when no voice is set. Select the API with providerOptions.azure.api to override model-based routing, and add azure.speechModel() as an alias of azure.speech().
e66dcd5: Reject an Azure resourceName that is not a single DNS label, so a malformed value cannot rewrite the request host.