Diffusion Models for Audio Quiz

Answers and explanations appear as you take the quiz.

1. What generative learning idea underlies the audio diffusion methods described here?
   • Only compressing every clip into MP 3
   • Only assigning a genre label to an unchanged file
   • Learning a reverse-corruption or denoising process in a specified representation
   • Only transcribing spoken words
2. Why can a studied latent audio method use a learned compressed representation?
   • Raw waveforms cannot be modeled by diffusion at all
   • It changes the modeled representation and can reduce its dimensional workload
   • Every latent guarantees perfect sound quality
   • Latent features contain no usable information
3. Which shared audio-language representation does the original AudioLDM study use for conditioning?
   • Only an arbitrary filename index
   • Only a mandatory image-pixel matrix
   • Only a MIDI notation grammar
   • CLAP embeddings
4. What role does a vocoder serve when used after generating mel-spectral features?
   • Converting the relevant spectral representation into waveform audio
   • Only printing the spectrogram on paper
   • Only renaming a generated file
   • Only adding a mandatory soundtrack to an image
5. Which related research direction learns a generation procedure that can use fewer sampling evaluations?
   • Only raising the sample rate without changing the model
   • Consistency learning or distillation under a tested setup
   • Only changing the output filename
   • A guarantee that disabling every condition preserves quality
6. Which cited audio research system uses timing-conditioned latent diffusion?
   • JPEG image compression
   • Bluetooth transport alone
   • HTTP document transfer alone
   • The original Stable Audio study
7. How does the original AudioLDM paper distinguish its diffusion conditioning during training and text-to-audio sampling?
   • Audio embeddings during training and text embeddings during sampling
   • Only text embeddings during all codec pretraining by definition
   • Only a filename during training and no conditions during sampling
   • A requirement to use the final output waveform as every future text prompt
8. Which cited model directly studies conditional and unconditional waveform diffusion generation?
   • A mandatory text-only dictionary model
   • An unchanged image caption index
   • DiffWave
   • A database transaction classifier