Spleeter Stem Separation Quiz

Answers and explanations appear as you take the quiz.

1. Which company originally released Spleeter?
   • Deezer
   • Spotify
   • Apple
   • Google
2. What does Spleeter's 4-stem model separate audio into?
   • Vocals, guitar, piano, strings
   • Vocals, drums, bass, other
   • Lead, harmony, melody, rhythm
   • Left, right, center, ambience
3. What neural network architecture does Spleeter use?
   • A recurrent LSTM on raw samples
   • A decision tree ensemble
   • A U-Net operating on spectrograms
   • A diffusion model
4. How does Spleeter actually extract a single source from the mix?
   • By subtracting a fixed template
   • By phase cancellation of two channels
   • By filtering with a fixed equalizer
   • By predicting a soft mask applied to the spectrogram
5. A common source of artifacts in Spleeter's output is that it:
   • Reuses the original mixture's phase for every stem
   • Uses lossy MP3 internally
   • Downsamples to 8 kHz
   • Runs only on CPU
6. What did Deezer report about the speed of Spleeter’s pretrained 4-stem model?
   • Real-time operation with no buffering on every device
   • About 100 times faster than real time on one GPU, with buffering still needed
   • A fixed speed of 100 times real time on a phone CPU
   • It can separate audio without processing the input
7. Which stem does Spleeter’s 5-stem configuration add to the 4-stem set?
   • Violin
   • Speech transcript
   • Piano
   • Room reverb
8. What do Spleeter’s pretrained U-Nets estimate for the target sources?
   • A text transcript of the song
   • A fixed equalizer curve for the whole track
   • A new recording of each instrument
   • Soft masks over a spectrogram representation