Prepare and test your skills

Prepare and test your skills

Free sample question with answers and detailed explanation below. The Cloud Digital Leader practice bank has an additional 3,070 practice questions.Try 10 free realistic questions: instant score, no credit card, and no account required.
A retail bank is upgrading its mobile application and Interactive Voice Response (IVR) system. The bank needs a solution that meets the following business requirements:
Which Google Cloud service should the bank implement?
Cloud Speech-to-Text API is an Automated Speech Recognition (ASR) service that uses neural network algorithms to convert spoken audio into written text transcriptions. Its streaming recognition mode processes live audio within a bidirectional stream using consecutive StreamingRecognizeRequest messages, returning real-time interim transcription results while a user is actively speaking into a microphone.
This service converts audio input into text output, which is the opposite of the bank's requirement to convert written text into playable audio. While streaming recognition is ideal for listening to a caller in real time, it cannot synthesize speech, read written account summaries aloud for visually impaired customers, or generate automated voice responses.
100-millisecond frame size to balance latency and efficiency).FLAC and LINEAR16 or lossy codecs like OGG_OPUS and MULAW, but cannot accept text transcripts for speech synthesis.The scenario asked for a solution to capture live spoken commands from customers over the phone in real time or generate live captions for hearing-impaired users during a live broadcast.
In a complete conversational voice agent or Interactive Voice Response (IVR) architecture, Cloud Speech-to-Text API with streaming recognition and Cloud Text-to-Speech API are frequently combined. Speech-to-Text first transcribes the customer's spoken request into text for a language model to process, and Text-to-Speech then reads the generated text response back to the caller.
Speech Adaptation is a customization capability within the Cloud Speech-to-Text API used to tune and improve the transcription accuracy of Automated Speech Recognition (ASR). It enables developers to supply word and phrase hints and custom class tokens to boost recognition accuracy for domain-specific vocabulary, proper names, and uncommon terms.
Speech Adaptation is designed to improve how accurately incoming spoken audio is recognized and transcribed into written text, helping to lower a system's Word Error Rate (WER). It does not convert written text into playable audio, support visually impaired users through screen-reading speech synthesis, or perform text normalization for pronouncing written dates and currency values.
The bank measured a high Word Error Rate (WER) when transcribing callers' spoken inquiries against human-provided ground truth transcripts because the recognizer failed to identify proprietary banking product names.
It is helpful to distinguish between Speech Adaptation in Speech-to-Text and text normalization in Text-to-Speech. Speech Adaptation helps a speech recognition model hear and transcribe specific spoken words into text accurately, whereas text normalization in Cloud Text-to-Speech API helps a speech synthesis model speak written dates, numbers, and monetary amounts naturally.
Cloud Speech-to-Text API with asynchronous recognition is a batch speech recognition method designed to transcribe long-form, pre-recorded audio files into written text. When audio data is sent to the API, it initiates a long-running operation that processes the audio in the background, allowing client applications to periodically poll for the completed text transcription.
First, like all Speech-to-Text methods, asynchronous recognition turns spoken audio into written text rather than synthesizing playable speech from written account summaries. Second, asynchronous processing relies on polling a background operation rather than streaming real-time audio chunks, making it unsuitable for generating interactive automated voice responses in an IVR system.
480 minutes (8 hours) in duration, far exceeding the 1-minute limit of synchronous recognition requests.FLAC or LINEAR16 recorded at 16,000 Hz (or native 8,000 Hz for legacy telephony audio).RecognizeResponse objects containing written transcripts rather than playable audio files.The bank needed to transcribe batches of recorded customer support phone calls up to 480 minutes long for offline quality assurance, compliance auditing, and searchable text archives.
Students sometimes confuse asynchronous recognition with automated voice generation because both are widely deployed in call center environments. However, asynchronous Speech-to-Text is an analytical backend workflow for converting recorded call audio into text, whereas Cloud Text-to-Speech API is a user-facing service for generating spoken audio from text.
Cloud Text-to-Speech API is a pre-trained Google Cloud artificial intelligence service that converts written text inputs into natural-sounding, synthetic human speech. By sending transcript data or Speech Synthesis Markup Language (SSML) in an API call, applications receive playable audio files or streams generated by advanced deep learning voice models such as Gemini-TTS and Chirp 3 HD.
audioConfig field.655 seconds of playable audio from up to 8,000 bytes of combined text and prompt input per request.Unlike Cloud Speech-to-Text API options—which process incoming audio to produce written transcripts—Cloud Text-to-Speech API performs the exact operation required here: synthesizing audio output from text. Furthermore, compared to calling speech synthesis models through the standard Vertex AI API (which outputs raw PCM 16-bit 24kHz audio data without WAV headers), Cloud Text-to-Speech API provides built-in text normalization, custom output audio encodings, and multi-chunk text streaming out of the box.
Answer them yourself and get your score the moment you finish. No account needed.
Full access includes
Keep the momentum going with these hand-picked practice scenarios
Last updated