From 7bf0984698e43b9aa2dd879f65951318ed410390 Mon Sep 17 00:00:00 2001
From: Maze <167211895+mazeincoding@users.noreply.github.com>
Date: Sat, 31 Jan 2026 00:20:04 +0100
Subject: codebase overhaul (#697)

---
 .../editor/panels/assets/views/captions.tsx        | 152 +++++++++++++++++++++
 1 file changed, 152 insertions(+)
 create mode 100644 apps/web/src/components/editor/panels/assets/views/captions.tsx

(limited to 'apps/web/src/components/editor/panels/assets/views/captions.tsx')

diff --git a/apps/web/src/components/editor/panels/assets/views/captions.tsx b/apps/web/src/components/editor/panels/assets/views/captions.tsx
new file mode 100644
index 00000000..306bad2a
--- /dev/null
+++ b/apps/web/src/components/editor/panels/assets/views/captions.tsx
@@ -0,0 +1,152 @@
+import { Button } from "@/components/ui/button";
+import { PropertyGroup } from "@/components/editor/panels/properties/property-item";
+import { PanelBaseView as BaseView } from "@/components/editor/panels/panel-base-view";
+import {
+	Select,
+	SelectContent,
+	SelectItem,
+	SelectTrigger,
+	SelectValue,
+} from "@/components/ui/select";
+import { useState, useRef } from "react";
+import { extractTimelineAudio } from "@/lib/media/mediabunny";
+import { useEditor } from "@/hooks/use-editor";
+import { DEFAULT_TEXT_ELEMENT } from "@/constants/text-constants";
+import { TRANSCRIPTION_LANGUAGES } from "@/constants/transcription-constants";
+import type {
+	TranscriptionLanguage,
+	TranscriptionProgress,
+} from "@/types/transcription";
+import { transcriptionService } from "@/services/transcription";
+import { decodeAudioToFloat32 } from "@/lib/media/audio";
+import { buildCaptionChunks } from "@/lib/transcription/caption";
+import { Spinner } from "@/components/ui/spinner";
+
+export function Captions() {
+	const [selectedLanguage, setSelectedLanguage] =
+		useState<TranscriptionLanguage>("auto");
+	const [isProcessing, setIsProcessing] = useState(false);
+	const [processingStep, setProcessingStep] = useState("");
+	const [error, setError] = useState<string | null>(null);
+	const containerRef = useRef<HTMLDivElement>(null);
+	const editor = useEditor();
+
+	const handleProgress = (progress: TranscriptionProgress) => {
+		if (progress.status === "loading-model") {
+			setProcessingStep(`Loading model ${Math.round(progress.progress)}%`);
+		} else if (progress.status === "transcribing") {
+			setProcessingStep("Transcribing...");
+		}
+	};
+
+	const handleGenerateTranscript = async () => {
+		try {
+			setIsProcessing(true);
+			setError(null);
+			setProcessingStep("Extracting audio...");
+
+			const audioBlob = await extractTimelineAudio({
+				tracks: editor.timeline.getTracks(),
+				mediaAssets: editor.media.getAssets(),
+				totalDuration: editor.timeline.getTotalDuration(),
+			});
+
+			setProcessingStep("Preparing audio...");
+			const { samples } = await decodeAudioToFloat32({ audioBlob });
+
+			const result = await transcriptionService.transcribe({
+				audioData: samples,
+				language: selectedLanguage === "auto" ? undefined : selectedLanguage,
+				onProgress: handleProgress,
+			});
+
+			setProcessingStep("Generating captions...");
+			const captionChunks = buildCaptionChunks({ segments: result.segments });
+
+			const captionTrackId = editor.timeline.addTrack({
+				type: "text",
+				index: 0,
+			});
+
+			for (let i = 0; i < captionChunks.length; i++) {
+				const caption = captionChunks[i];
+				editor.timeline.insertElement({
+					placement: { mode: "explicit", trackId: captionTrackId },
+					element: {
+						...DEFAULT_TEXT_ELEMENT,
+						name: `Caption ${i + 1}`,
+						content: caption.text,
+						duration: caption.duration,
+						startTime: caption.startTime,
+						fontSize: 65,
+						fontWeight: "bold",
+					},
+				});
+			}
+		} catch (error) {
+			console.error("Transcription failed:", error);
+			setError(
+				error instanceof Error ? error.message : "An unexpected error occurred",
+			);
+		} finally {
+			setIsProcessing(false);
+			setProcessingStep("");
+		}
+	};
+
+	const handleLanguageChange = ({ value }: { value: string }) => {
+		if (value === "auto") {
+			setSelectedLanguage("auto");
+			return;
+		}
+
+		const matchedLanguage = TRANSCRIPTION_LANGUAGES.find(
+			(language) => language.code === value,
+		);
+		if (!matchedLanguage) return;
+		setSelectedLanguage(matchedLanguage.code);
+	};
+
+	return (
+		<BaseView
+			ref={containerRef}
+			className="flex h-full flex-col justify-between"
+		>
+			<PropertyGroup title="Language">
+				<Select
+					value={selectedLanguage}
+					onValueChange={(value) => handleLanguageChange({ value })}
+				>
+					<SelectTrigger className="bg-panel-accent h-8 w-full text-xs">
+						<SelectValue placeholder="Select a language" />
+					</SelectTrigger>
+					<SelectContent>
+						<SelectItem value="auto">Auto detect</SelectItem>
+						{TRANSCRIPTION_LANGUAGES.map((language) => (
+							<SelectItem key={language.code} value={language.code}>
+								{language.name}
+							</SelectItem>
+						))}
+					</SelectContent>
+				</Select>
+			</PropertyGroup>
+
+			<div className="flex flex-col gap-4">
+				{error && (
+					<div className="bg-destructive/10 border-destructive/20 rounded-md border p-3">
+						<p className="text-destructive text-sm">{error}</p>
+					</div>
+				)}
+
+				<Button
+					className="w-full"
+					onClick={handleGenerateTranscript}
+					disabled={isProcessing}
+				>
+					{isProcessing && <Spinner className="mr-1" />}
+					{isProcessing ? processingStep : "Generate transcript"}
+				</Button>
+			</div>
+		</BaseView>
+	);
+}
-- 
cgit v1.3.1