Voice Conversion Quiz

Answers and explanations appear as you take the quiz.

1. What does voice conversion change about a recording?
   • The identity of the speaker, who it sounds like
   • The words that are spoken
   • The language of the content
   • The background music only
2. What core capability lets a system swap a voice while keeping the words?
   • Increasing the audio bitrate
   • Disentangling linguistic content from speaker identity
   • Translating the text first
   • Removing all pitch from the signal
3. How does AutoVC encourage the model to strip out speaker identity from content?
   • By using a very large unrestricted hidden layer
   • By requiring thousands of parallel recordings
   • By using a carefully sized information bottleneck
   • By adding random noise to the words
4. What distinguishes a 'parallel' voice conversion dataset from a 'non-parallel' one?
   • Parallel data is always recorded in stereo
   • Non-parallel data must be in two languages
   • There is no real difference between them
   • Parallel data requires both speakers to say the same sentences; non-parallel does not
5. What makes the AutoVC approach “zero-shot” in the paper’s voice-conversion setting?
   • It converts to a target speaker not seen as a training speaker, using target reference audio without retraining for that target
   • It produces speech without any audio input
   • It translates speech into a new language automatically
   • It requires paired recordings of both speakers saying every sentence
6. Why does ASVspoof evaluate spoofed or deepfake speech countermeasures?
   • To measure whether a voice converter preserves punctuation
   • To test defenses that protect automatic speaker verification from manipulated audio
   • To identify the language of every sentence
   • To replace the enrollment utterance with a text transcript
7. How does CycleGAN-VC learn a mapping between speakers without parallel utterances?
   • Exact word-for-word paired recordings are mandatory
   • It translates only the written transcript
   • Cycle-consistency and adversarial objectives constrain conversions between domains
   • It copies the target speaker’s raw waveform
8. In a voice-conversion system, what does a vocoder do?
   • Identifies the speaker’s legal identity
   • Translates speech into another language
   • Measures speaker-verification accuracy
   • Converts acoustic features such as a spectrogram into an audio waveform