A paired audit of three TTS systems shows that descriptor-aligned voice changes come with off-target acoustic shifts, and a candidate selector reduces these shifts at inference time.
Policies and operating point The candidate budget counts descriptor-conditioned generations; the matched neutral baseline y0 is shared across all policies and is not included in B
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
other 1
citation-polarity summary
fields
cs.SD 1years
2026 1verdicts
CONDITIONAL 1roles
other 1polarities
unclear 1representative citing papers
citing papers explorer
-
Beyond Prompt Adherence: Auditing Attribute-Level Voice Control in Speech Generation
A paired audit of three TTS systems shows that descriptor-aligned voice changes come with off-target acoustic shifts, and a candidate selector reduces these shifts at inference time.