33 lines
1.4 KiB
TypeScript
33 lines
1.4 KiB
TypeScript
|
// Dev sanity check for the audio segmenter. Run: npx tsx scripts/test-segment.ts
|
||
|
|
import { segmentTurns, splitLongText, flattenTurns } from "@/lib/ai/pipeline/segment";
|
||
|
|
import type { StructuredScript } from "@/lib/ai/types";
|
||
|
|
|
||
|
|
const turns = [
|
||
|
|
{ text: "Hello and welcome to the show.", voiceId: "A" },
|
||
|
|
{ text: "x".repeat(2500), voiceId: "B" }, // oversized single turn
|
||
|
|
{ text: "Thanks for having me.", voiceId: "A" },
|
||
|
|
{ text: "Let's dive in.", voiceId: "A" },
|
||
|
|
];
|
||
|
|
|
||
|
|
const segs = segmentTurns(turns, 1800);
|
||
|
|
console.log(`segmentTurns -> ${segs.length} segments`);
|
||
|
|
segs.forEach((s, i) =>
|
||
|
|
console.log(` #${i}: chars=${s.characters} voices=${s.uniqueVoices} turns=${s.turns.length}`)
|
||
|
|
);
|
||
|
|
|
||
|
|
const parts = splitLongText("One. Two. Three. " + "w".repeat(40), 15);
|
||
|
|
console.log(`splitLongText -> ${parts.length} parts (max len ${Math.max(...parts.map((p) => p.length))})`);
|
||
|
|
|
||
|
|
const script: StructuredScript = {
|
||
|
|
title: "T",
|
||
|
|
sections: [
|
||
|
|
{ id: "a", title: "A", turns: [{ speakerKey: "host", text: "Hi" }, { speakerKey: "ghost", text: "" }] },
|
||
|
|
{ id: "b", title: "B", turns: [{ speakerKey: "guest", text: "Hello" }] },
|
||
|
|
],
|
||
|
|
};
|
||
|
|
const flat = flattenTurns(script, { host: "V1", guest: "V2" }, "V1");
|
||
|
|
console.log(`flattenTurns -> ${flat.length} turns (empty dropped), voices: ${flat.map((t) => t.voiceId).join(",")}`);
|
||
|
|
|
||
|
|
const allUnder = segs.every((s) => s.characters <= 1800);
|
||
|
|
console.log(allUnder ? "PASS: all segments within limit" : "FAIL: a segment exceeds the limit");
|