Public derived data package for Acoustic Deviations between AI-Synthesized and Original Human Singing

Description

This dataset contains anonymized derived acoustic feature tables, sanitized Praat script templates, statistical summary scripts, and figure-source tables for a same-song paired study of AI-synthesized or converted Cantonese singing and original human singing. Full commercial audio, complete AI-cover audio, and full lyric-bearing TextGrid files are excluded because of copyright, lyric-rights, and voice-model use constraints.

Steps to reproduce

The public package was generated from the experiment materials by retaining only derived acoustic measurements and scripts. Praat track tables, segment interval tables, burst/release event tables, peak measurements, and figure-source summaries are anonymized by pair and excerpt identifiers. Run scripts/statistical_summary/summarize_public_tables.py from the package root to reproduce key public summaries.

Institutions

• Guangzhou University

  Guangdong, Guangzhou

Categories

Funders

• Guangzhou University

  Guangzhou

Licence