const BENCHMARK_COPY = {
text: {
title: "Text Editing",
description:
"Localized substitution, insertion, deletion, and longer text edits in English and Mandarin.",
},
emotion: {
title: "Emotion Editing",
description:
"Phrase-level, multi-span, and whole-utterance emotion control.",
},
prosody: {
title: "Prosody Editing",
description:
"Pitch and speaking-rate control over one or more selected regions.",
},
pause: {
title: "Pause Editing",
description:
"Insertion and reduction of natural pauses at one or more boundaries.",
},
multi: {
title: "Compositional Editing",
description:
"Up to three text and acoustic operations in one instruction.",
},
noise: {
title: "Noise Preservation & Denoising",
description:
"Editing speech while preserving background sound, plus enhancement for denoising.",
},
multi_speaker: {
title: "Multi-speaker Preservation",
description:
"Editing a bilingual, two-speaker source without speaker labels in the instruction.",
},
};
const escapeHtml = (value) =>
String(value)
.replaceAll("&", "&")
.replaceAll("<", "<")
.replaceAll(">", ">")
.replaceAll('"', """);
function audioCell(label, path, kind = "source") {
const isOurs = kind === "ours";
return `