GCP Professional Cloud Architect Practice Question

A pre-trained, fully managed service is required that can accept long audio files from Cloud Storage, perform asynchronous processing, add automatic punctuation, and return word-level timestamps. Cloud Speech-to-Text meets all of these needs through its long-running recognise methods and optional features such as automatic punctuation and word time offsets. Cloud Text-to-Speech performs the opposite task-generating audio from text. Transcoder API handles video and audio format conversion but does not create transcripts. Building and hosting a custom model on Vertex AI would contradict the requirement to avoid managing speech-recognition infrastructure.