Users praise audio LLMs that generate demonic speech by optimizing noise input because the results sound originally made rather than mimicking human voices.
Based on 1 visible X reactions from 5 accounts; directional sample.
Ask a question below.
Published answers will appear here.
Optimizing raw noise input reveals how they implicitly encode voices.
@mayfer I'd pick the demon voice too. Sounds *made*, not mimicked. Voice LLMs chase human likeness. Deepdream chases feature activation. Uncanny valley, inverted.
ok so first of all this is amazing and i'm so glad matt discovered it, i can even understand it when i stop reading the transcript second of all why do i prefer the way deepdream speaks to real voice generating llm's
Users praise audio LLMs that generate demonic speech by optimizing noise input because the results sound originally made rather than mimicking human voices.
Based on 1 visible X reactions from 5 accounts; directional sample.
Ask a question below.
Published answers will appear here.