resgen-genai

Audio Samples

Uncurated Samples from ResGen (625M) Models

This section provides examples of uncurated audio samples generated by the ResGen (625M) models. The samples demonstrate the model’s performance in two tasks:

  1. Samples of Continuation Task:
    • Generated Sample Details: The first 3 seconds are ground truth speech, and the subsequent portion is generated by the model.
  2. Samples of Cross-Sentence Task:
    • Generated Sample Details: The italicized portion of the text is used as the prompt for the model, and the remaining portion of the audio is generated.
Continuation Task Cross Task
Text: “he began a confused complaint against the wizard who had vanished behind the curtain on the left” Prompt:he looked at me but as if instead of
Text: “housecleaning a domestic upheaval that makes it easy for the government to enlist all the soldiers it needs”
Text: “before them fled the stroller and his three sons capless and terrified” Prompt:the tiresome product of a tireless tongue
Text: “there was a unanimous groan at this and much reproach after which in his preoccupied way he explained”
Text: “i could not see my boy injured excellence for but doing his duty as one of cumberland’s sons” Prompt:makes it easy for the government to enlist
Text: “well if i don’t know who she was in love with i know who he was”
Text: “we will go out together to the bower there is a way down to the court from my window” Prompt:strange and all the more so because of his
Text: “there were plenty of people to help but of course the young lady who should go down as governess would be in supreme authority”
Text: “they regained their apartment apparently without disturbing the household of gamewell” Prompt:postponement but it was just his scruples that charmed
Text: “it sounded dull it sounded strange and all the more so because of his main condition which was”