jaimemedina/emoscale-intensity-reference-pilot

A 12-group listening diagnostic. Use headphones, fixed comfortable volume, and the invitation code provided by Jaime. Complete only your assigned form. Responses stay in your browser: download the JSON and send it directly to Jaime. There is no automatic response upload. Invitations assign forms; they do not make the hosted audio private. Please do not inspect site assets before rating.

Research scope and attribution — read after completing the study

This exploratory diagnostic includes audio from EmoScale and the demonstration accompanying Masato Murata, Koichi Miyazaki and Tomoki Koriyama, "Speaker-agnostic Emotion Vector for Cross-speaker Emotion Intensity Control", Interspeech 2025. Paper · Original demonstration

Original demonstrations are reproduced alongside constant-gain loudness-matched copies for this research comparison. Other stimuli are EmoScale outputs derived from ESD. Third-party audio remains attributable to its original authors; this Space does not grant a new license to the source recordings. The supplied demo repository does not contain an explicit audio redistribution license.

System, emotion, intensity, source and processing labels are not attached to the listening trials. The answer key and plaintext invitation codes are not uploaded. Different texts, voices, synthesis tasks and intensity ranges mean this is NOT a matched system benchmark or a replication of the paper's listener study.