Bark Generative Audio Model Quiz

Answers and explanations appear as you take the quiz.

1. What makes Bark different from a traditional text-to-speech engine?
   • It can generate nonverbal sounds like laughter and music, not just speech
   • It only outputs clean robotic speech
   • It requires a recorded voice sample for every output
   • It converts audio into text
2. Who released the Bark model?
   • OpenAI
   • Suno
   • Google DeepMind
   • Meta
3. How does Bark fundamentally generate audio?
   • By stitching together pre-recorded phonemes
   • By applying filters to a reference recording
   • By predicting discrete audio tokens autoregressively like a language model
   • By rendering MIDI notes
4. What neural codec does Bark use to turn tokens into a waveform?
   • MP3
   • FLAC
   • WaveNet
   • Meta's EnCodec
5. Why might the same Bark prompt produce different results each time?
   • Generation is probabilistic, so each run samples a different sequence
   • It caches outputs randomly
   • It always errors on repeat prompts
   • It depends on the time of day
6. What is a known drawback of Bark compared to dedicated TTS systems?
   • It cannot speak any language
   • It can hallucinate extra sounds and is less controllable
   • It only works offline
   • It produces silent audio
7. What does Bark’s semantic-to-coarse stage produce?
   • A transcript of the generated waveform
   • A mel-spectrogram with no audio tokens
   • The first two EnCodec audio codebooks
   • An image embedding
8. How can a Bark prompt request a nonverbal event such as laughter?
   • Add an image mask to the audio tokens
   • Train a new EnCodec model for every cue
   • Convert the cue to a CSV file
   • Include a cue such as [laughs] in the text prompt