mirror of
https://github.com/webadderallorg/Recordly.git
synced 2026-11-09 15:33:31 +00:00
47 lines
1.8 KiB
TypeScript
47 lines
1.8 KiB
TypeScript
import { expect, it } from "vitest";
|
||
import { buildCaptionTextFromWords, parseWhisperJsonWords } from "./parser";
|
||
|
||
const timed = (text: string, from: number, to: number) => ({ text, offsets: { from, to } });
|
||
it("retains token-aligned word boundaries in Chinese without introducing spaces or times", () => {
|
||
const words = parseWhisperJsonWords([
|
||
timed("你", 0, 100),
|
||
timed("好", 100, 180),
|
||
timed("世界", 200, 500),
|
||
timed("！", 0, 0),
|
||
]);
|
||
expect(words).toEqual([
|
||
{ text: "你好", startMs: 0, endMs: 180 },
|
||
{ text: "世界！", startMs: 200, endMs: 500 },
|
||
]);
|
||
expect(buildCaptionTextFromWords(words)).toBe("你好世界！");
|
||
});
|
||
it("preserves a model token containing multiple words rather than fabricating alignments", () => {
|
||
const words = parseWhisperJsonWords([timed("你好世界", 50, 600)]);
|
||
expect(words).toEqual([{ text: "你好世界", startMs: 50, endMs: 600 }]);
|
||
});
|
||
it("handles Japanese token boundaries and mixed Latin text without adding whitespace", () => {
|
||
const words = parseWhisperJsonWords([
|
||
timed("私", 0, 100),
|
||
timed("は", 100, 200),
|
||
timed("東京", 200, 500),
|
||
timed("です", 500, 700),
|
||
]);
|
||
expect(words.map((word) => word.text)).toEqual(["私", "は", "東京", "です"]);
|
||
expect(buildCaptionTextFromWords(words)).toBe("私は東京です");
|
||
expect(words.map((word) => word.startMs)).toEqual([0, 100, 200, 500]);
|
||
});
|
||
it("keeps English subwords, punctuation and real spaces unchanged", () => {
|
||
const words = parseWhisperJsonWords([
|
||
timed("[_BEG_]", 0, 0),
|
||
timed("Hel", 0, 120),
|
||
timed("lo", 120, 220),
|
||
timed(",", 0, 0),
|
||
timed(" world", 250, 600),
|
||
timed("!", 0, 0),
|
||
]);
|
||
expect(words).toEqual([
|
||
{ text: "Hello,", startMs: 0, endMs: 220 },
|
||
{ text: "world!", startMs: 250, endMs: 600, leadingSpace: true },
|
||
]);
|
||
});
|