AI-900 Practice Question: Describe features of generative AI workloads on Azure

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✓

  Computer Vision: Image (file or URL)

  Why this is correct

  The Computer Vision service accepts a binary image file or a publicly accessible image URL as its primary input. It uses convolutional neural networks to analyze pixel-level data for tasks such as object detection, image classification, and optical character recognition (OCR). Supplying a URL or file is required so the service can fetch, decode, and process the visual content.

• ✓

  Text Analytics: Text documents

  Why this is correct

  Text Analytics, now part of Azure AI Language, expects raw unstructured text documents as input, typically passed as strings in a JSON request payload. It performs language detection, sentiment analysis, key phrase extraction, and named entity recognition by parsing the linguistic structure and semantics of the text. Unlike vision or speech services, it operates purely on written language and does not consume image pixels or audio waveforms.

• ✓

  Speech: Audio files or streams

  Why this is correct

  The Speech service consumes audio files, such as WAV or MP3, or real-time audio streams captured from a microphone or telephony system. Its acoustic models analyze the waveform, sample rate, and encoding to perform speech-to-text transcription, speech translation, or speaker recognition. The input must be a time-series audio signal rather than a static text document or image, because the service is designed to interpret spoken language.

• ✗

  Computer Vision: Text documents

  Why it's wrong here

  Feeding text documents to Computer Vision is incorrect because the service only accepts image data, such as JPEG, PNG, or TIFF files, or image URLs, and all of its algorithms rely on visual structure. Although Computer Vision can perform OCR to extract text from a picture, the input remains an image, not a .txt or .docx file. A plain text document contains no pixel grid or visual features, so the image-analysis endpoint would reject it.

• ✗

  Speech: Image (file or URL)

  Why it's wrong here

  Images are not a valid input for the Speech service because the service's recognition and synthesis pipelines are built around audio signals, not visual content. An image file or URL contains no waveform, sample rate, channel count, or audio encoding that the speech models can process. Image inputs are intended for Computer Vision, so sending a JPG to a Speech endpoint would trigger an unsupported media type error rather than producing a transcription.

Go deeper

Related to this question