Record your screen, ship a demo. Free and open-source, GPU-accelerated, no watermarks, no subscriptions. Windows, macOS, Linux. Actively maintained.
TypeScript cross-platformelectronffmpegopen-sourceproduct-demorustscreen-capturescreen-recorderscreen-recordingscreen-studioscreencastspeech-to-texttypescriptvideo-editingvideo-editorwaylandwgpuwhisper
// The caption tools end to end on a document: turn them on, style and place them,
// translate them as a client would, and check the export draws what was asked.
import { describe, expect, it } from "vitest";
import { CAPTION_STYLES, getCaptionSettings } from "../../src/lib/ai-edition/captions";
import {
type AxcutDocument,
createEmptyDocument,
documentSchema,
} from "../../src/lib/ai-edition/schema";
import { buildSceneDescription } from "../../src/native/sceneDescription";
import { type AgentToolExecution, executeAgentTool } from "./agent-tools";
function fixtureDocument(): AxcutDocument {
const base = createEmptyDocument({
title: "Captions",
projectId: "proj_1",
createdAt: "2026-01-01T00:00:00.000Z",
});
const words = [
["w1", "Hello", 0, 0.4],
["w2", "there.", 0.4, 0.8],
["w3", "Second", 2, 2.4],
["w4", "phrase.", 2.4, 2.8],
] as const;
return documentSchema.parse({
...base,
project: { ...base.project, primaryAssetId: "asset_1" },
assets: [
{ id: "asset_1", kind: "video", label: "Rec", originalPath: "/tmp/rec.mp4", durationSec: 10 },
],
// One segment per word, as Whisper's transcripts are stored.
transcripts: [
{
assetId: "asset_1",
language: "en",
segments: words.map(([id, text, startSec, endSec]) => ({
id: `seg_${id}`,
kind: "speech",
startSec,
endSec,
text,
wordIds: [id],
})),
words: words.map(([id, text, startSec, endSec]) => ({
id,
segmentId: `seg_${id}`,
startSec,
endSec,
text,
})),
},
],
timeline: {
...base.timeline,
clips: [
{
id: "clip_1",
assetId: "asset_1",
sourceStartSec: 0,
sourceEndSec: 10,
timelineStartSec: 0,
timelineEndSec: 10,
wordRefs: [],
origin: "user",
reason: "",
},
],
},
});
}
function call(document: AxcutDocument, name: string, args: unknown = {}) {
const execution: AgentToolExecution = executeAgentTool(document, name, JSON.stringify(args));
return { ...execution, result: JSON.parse(execution.resultJson) as Record<string, unknown> };
}
function captionTexts(document: AxcutDocument): string[] {
return buildSceneDescription(document).annotations.map((a) => a.text?.content ?? "");
}
describe("setCaptionSettings", () => {
it("turns the captions on, and the export draws the transcript", () => {
const doc = fixtureDocument();
expect(captionTexts(doc)).toEqual([]);
const on = call(doc, "setCaptionSettings", { enabled: true });
expect(on.ok).toBe(true);
expect(on.document && captionTexts(on.document).join(" ")).toContain("Hello there.");
});
it("applies a style, then a plate, then the explicit fields over both", () => {
const doc = fixtureDocument();
const set = call(doc, "setCaptionSettings", {
style: "bold",
plate: "light",
fontSize: 72,
anchorV: "top",
insetY: 8,
minWordsPerLine: 3,
maxWordsPerLine: 5,
});
const settings = getCaptionSettings(set.document);
expect(settings.fontFamily).toBe(CAPTION_STYLES.bold.fontFamily);
expect(settings.fontSize).toBe(72);
expect(settings.backgroundColor).toBe("#ffffff");
// The bold style's yellow would vanish on a light plate: the plate recoloured it.
expect(settings.color).not.toBe(CAPTION_STYLES.bold.color);
expect(settings).toMatchObject({ anchorV: "top", insetY: 8, minWordsPerLine: 3 });
expect(set.result.plate).toBe("light");
});
it("refuses what the pane cannot express, and changes nothing", () => {
const doc = fixtureDocument();
expect(call(doc, "setCaptionSettings", {}).ok).toBe(false);
expect(call(doc, "setCaptionSettings", { minWordsPerLine: 9, maxWordsPerLine: 3 }).ok).toBe(
false,
);
expect(call(doc, "setCaptionSettings", { fontSize: 500 }).ok).toBe(false);
expect(call(doc, "setCaptionSettings", { fontFamily: "Comic Sans" }).ok).toBe(false);
const missing = call(doc, "setCaptionSettings", { language: "fr" });
expect(missing.ok).toBe(false);
expect(String(missing.result.error)).toContain("putCaptionTranslation");
});
it("reports the lane the captions really read", () => {
const set = call(fixtureDocument(), "setCaptionSettings", { lane: "voiceover" });
expect(set.result.resolvedLane).toBe("recording");
expect(set.result.laneNote).toBeDefined();
});
});
describe("getCaptions", () => {
it("reads the settings back, warns when there is nothing to caption", () => {
const doc = call(fixtureDocument(), "setCaptionSettings", { enabled: true }).document;
if (!doc) throw new Error("not applied");
const read = call(doc, "getCaptions").result;
expect(read).toMatchObject({ style: "classic", plate: "dark", resolvedLane: "recording" });
expect(read.warning).toBeUndefined();
const silent = call({ ...doc, transcripts: [], transcript: null }, "getCaptions").result;
expect(silent.warning).toBeDefined();
});
});
describe("translating the captions", () => {
it("reads phrase units, stores the client's translation, and the export shows it", () => {
let doc = fixtureDocument();
const units = call(doc, "getCaptionTranslationUnits", { language: "fr" }).result
.assets as Array<{
units: Array<{ id: string; text: string }>;
}>;
// Phrases, not words: one unit per sentence.
expect(units[0]?.units.map((u) => u.text)).toEqual(["Hello there.", "Second phrase."]);
const [first, second] = units[0]?.units ?? [];
const put = call(doc, "putCaptionTranslation", {
language: "fr",
label: "Français",
assetId: "asset_1",
units: [{ id: first.id, text: "Bonjour." }],
});
expect(put.result).toMatchObject({
written: 1,
coverage: { translated: 1, total: 2 },
shown: false,
});
doc = put.document ?? doc;
// Only what is left to translate comes back.
const left = call(doc, "getCaptionTranslationUnits", { language: "fr" }).result
.assets as Array<{
units: Array<{ id: string }>;
}>;
expect(left[0]?.units.map((u) => u.id)).toEqual([second.id]);
doc = call(doc, "setCaptionSettings", { enabled: true, language: "fr" }).document ?? doc;
expect(captionTexts(doc).join(" ")).toContain("Bonjour.");
// The transcript itself is untouched.
expect(doc.transcripts[0]?.words[0]?.text).toBe("Hello");
const read = call(doc, "getCaptions").result.translations as Array<Record<string, unknown>>;
expect(read).toEqual([
expect.objectContaining({
language: "fr",
label: "Français",
coverage: [{ assetId: "asset_1", translated: 1, total: 2 }],
}),
]);
});
it("refuses unit ids it does not know, and nothing is written", () => {
const put = call(fixtureDocument(), "putCaptionTranslation", {
language: "fr",
assetId: "asset_1",
units: [{ id: "seg_w1", text: "Bonjour" }],
});
expect(put.ok).toBe(false);
expect(put.document).toBeUndefined();
const twice = call(fixtureDocument(), "putCaptionTranslation", {
language: "fr",
assetId: "asset_1",
units: [
{ id: "u:seg_w1", text: "Bonjour." },
{ id: "u:seg_w1", text: "Salut." },
],
});
expect(twice.ok).toBe(false);
});
it("removing the language on screen puts the original back", () => {
let doc = fixtureDocument();
doc =
call(doc, "putCaptionTranslation", {
language: "fr",
assetId: "asset_1",
units: [{ id: "u:seg_w1", text: "Bonjour." }],
}).document ?? doc;
doc = call(doc, "setCaptionSettings", { language: "fr" }).document ?? doc;
const removed = call(doc, "removeCaptionTranslation", { language: "fr" });
expect(removed.result.shownLanguage).toBeNull();
expect(call(removed.document ?? doc, "removeCaptionTranslation", { language: "fr" }).ok).toBe(
false,
);
});
});
describe("setVocabulary", () => {
it("stores the names Whisper is primed with, and getCurrentDocument reads them", () => {
const set = call(fixtureDocument(), "setVocabulary", { vocabulary: " OpenScreen, Vulkan " });
expect(set.document?.project.vocabulary).toBe("OpenScreen, Vulkan");
const read = set.document && call(set.document, "getCurrentDocument").result;
expect(read?.vocabulary).toBe("OpenScreen, Vulkan");
expect(call(fixtureDocument(), "setVocabulary", { vocabulary: "x".repeat(601) }).ok).toBe(
false,
);
});
});