EnCodec Audio Compression Quiz

Answers and explanations appear as you take the quiz.

1. Which organization released EnCodec?
   • Meta AI
   • OpenAI
   • Microsoft Research
   • NVIDIA
2. What extra component does EnCodec add to losslessly squeeze more compression from its tokens?
   • A spectrogram inverter
   • A Transformer-based entropy (language) model
   • A noise gate
   • A second decoder
3. At roughly 6 kbps, EnCodec's quality was reported as comparable to MP3 at about what bitrate?
   • 320 kbps
   • 16 kbps
   • 64 kbps
   • 8 kbps
4. What problem does EnCodec's 'balancer' solve during training?
   • It removes silence from recordings
   • It converts mono audio to stereo
   • It selects the sampling rate
   • It automatically scales competing loss terms so training stays stable
5. Which core quantization method does EnCodec share with SoundStream?
   • Residual vector quantization (RVQ)
   • k-means clustering only
   • Huffman coding
   • Principal component analysis
6. Which generative models use EnCodec tokens as their audio representation?
   • DALL-E and Midjourney
   • MusicGen and AudioGen
   • Whisper and Wav2Vec
   • Stable Diffusion and Imagen
7. Which audio formats are described for EnCodec in its paper?
   • 8 kHz mono only
   • 44.1 kHz stereo only
   • 24 kHz monophonic audio and 48 kHz stereophonic audio
   • 96 kHz multichannel audio only
8. Which losses help EnCodec balance waveform reconstruction and perceptual sound quality?
   • Only a text-token cross-entropy loss
   • Only a speaker-classification loss
   • A reward-model loss without reconstruction
   • Time-domain and multi-scale spectrogram reconstruction losses with adversarial discriminators