From 5b059ce6deb2b4132b445a0db3816388367acccf Mon Sep 17 00:00:00 2001
From: Maze Winther <mazewinther@gmail.com>
Date: Mon, 11 Aug 2025 22:24:54 +0200
Subject: feat: auto-captions

---
 .../editor/media-panel/views/base-view.tsx         |  16 +-
 .../editor/media-panel/views/captions.tsx          | 322 ++++++++++++++++++++-
 2 files changed, 325 insertions(+), 13 deletions(-)

(limited to 'apps/web/src/components/editor/media-panel')

diff --git a/apps/web/src/components/editor/media-panel/views/base-view.tsx b/apps/web/src/components/editor/media-panel/views/base-view.tsx
index 296295fc..2cd33535 100644
--- a/apps/web/src/components/editor/media-panel/views/base-view.tsx
+++ b/apps/web/src/components/editor/media-panel/views/base-view.tsx
@@ -11,12 +11,19 @@ interface BaseViewProps {
     content: React.ReactNode;
   }[];
   className?: string;
+  ref?: React.RefObject<HTMLDivElement>;
 }
 
-function ViewContent({ children }: { children: React.ReactNode }) {
+function ViewContent({
+  children,
+  className,
+}: {
+  children: React.ReactNode;
+  className?: string;
+}) {
   return (
     <ScrollArea className="flex-1">
-      <div className="p-5">{children}</div>
+      <div className={`p-5 h-full ${className}`}>{children}</div>
     </ScrollArea>
   );
 }
@@ -26,11 +33,12 @@ export function BaseView({
   defaultTab,
   tabs,
   className = "",
+  ref,
 }: BaseViewProps) {
   return (
-    <div className={`h-full flex flex-col ${className}`}>
+    <div className={`h-full flex flex-col ${className}`} ref={ref}>
       {!tabs || tabs.length === 0 ? (
-        <ViewContent>{children}</ViewContent>
+        <ViewContent className={className}>{children}</ViewContent>
       ) : (
         <Tabs defaultValue={defaultTab} className="flex flex-col h-full">
           <div className="px-3 pt-4 pb-0">
diff --git a/apps/web/src/components/editor/media-panel/views/captions.tsx b/apps/web/src/components/editor/media-panel/views/captions.tsx
index ef919fec..4fe2cbd7 100644
--- a/apps/web/src/components/editor/media-panel/views/captions.tsx
+++ b/apps/web/src/components/editor/media-panel/views/captions.tsx
@@ -1,9 +1,313 @@
-import { BaseView } from "./base-view";
-
-export function Captions() {
-  return (
-    <BaseView>
-      <div>Captions</div>
-    </BaseView>
-  );
-}
+import { Button } from "@/components/ui/button";
+import { PropertyGroup } from "../../properties-panel/property-item";
+import { BaseView } from "./base-view";
+import { Language, LanguageSelect } from "@/components/language-select";
+import { useState, useRef, useEffect } from "react";
+import { extractTimelineAudio } from "@/lib/ffmpeg-utils";
+import { encryptWithRandomKey, arrayBufferToBase64 } from "@/lib/zk-encryption";
+import { useTimelineStore } from "@/stores/timeline-store";
+import { Loader2, Shield, Trash2, Upload } from "lucide-react";
+import {
+  Dialog,
+  DialogContent,
+  DialogDescription,
+  DialogFooter,
+  DialogHeader,
+  DialogTitle,
+} from "@/components/ui/dialog";
+import { TextElement } from "@/types/timeline";
+
+export const languages: Language[] = [
+  { code: "US", name: "English" },
+  { code: "ES", name: "Spanish" },
+  { code: "IT", name: "Italian" },
+  { code: "FR", name: "French" },
+  { code: "DE", name: "German" },
+  { code: "PT", name: "Portuguese" },
+  { code: "RU", name: "Russian" },
+  { code: "JP", name: "Japanese" },
+  { code: "CN", name: "Chinese" },
+];
+
+const PRIVACY_DIALOG_KEY = "opencut-transcription-privacy-accepted";
+
+export function Captions() {
+  const [selectedCountry, setSelectedCountry] = useState("auto");
+  const [isProcessing, setIsProcessing] = useState(false);
+  const [processingStep, setProcessingStep] = useState<string>("");
+  const [error, setError] = useState<string | null>(null);
+  const [showPrivacyDialog, setShowPrivacyDialog] = useState(false);
+  const [hasAcceptedPrivacy, setHasAcceptedPrivacy] = useState(false);
+  const containerRef = useRef<HTMLDivElement>(null);
+  const { insertTrackAt, addElementToTrack } = useTimelineStore();
+
+  // Check if user has already accepted privacy on mount
+  useEffect(() => {
+    const hasAccepted = localStorage.getItem(PRIVACY_DIALOG_KEY) === "true";
+    setHasAcceptedPrivacy(hasAccepted);
+  }, []);
+
+  const handleGenerateTranscript = async () => {
+    try {
+      setIsProcessing(true);
+      setError(null);
+      setProcessingStep("Extracting audio...");
+
+      const audioBlob = await extractTimelineAudio();
+
+      setProcessingStep("Encrypting audio...");
+
+      // Encrypt the audio with a random key (zero-knowledge)
+      const audioBuffer = await audioBlob.arrayBuffer();
+      const encryptionResult = await encryptWithRandomKey(audioBuffer);
+
+      // Convert encrypted data to blob for upload
+      const encryptedBlob = new Blob([encryptionResult.encryptedData]);
+
+      setProcessingStep("Uploading...");
+      const uploadResponse = await fetch("/api/get-upload-url", {
+        method: "POST",
+        headers: { "Content-Type": "application/json" },
+        body: JSON.stringify({ fileExtension: "wav" }),
+      });
+
+      if (!uploadResponse.ok) {
+        const error = await uploadResponse.json();
+        throw new Error(error.message || "Failed to get upload URL");
+      }
+
+      const { uploadUrl, fileName } = await uploadResponse.json();
+
+      // Upload to R2
+      await fetch(uploadUrl, {
+        method: "PUT",
+        body: encryptedBlob,
+      });
+
+      setProcessingStep("Transcribing...");
+
+      // Call Modal transcription API with encryption parameters
+      const transcriptionResponse = await fetch("/api/transcribe", {
+        method: "POST",
+        headers: { "Content-Type": "application/json" },
+        body: JSON.stringify({
+          filename: fileName,
+          language:
+            selectedCountry === "auto" ? "auto" : selectedCountry.toLowerCase(),
+          // Send the raw encryption key and IV (zero-knowledge)
+          decryptionKey: arrayBufferToBase64(encryptionResult.key),
+          iv: arrayBufferToBase64(encryptionResult.iv),
+        }),
+      });
+
+      if (!transcriptionResponse.ok) {
+        const error = await transcriptionResponse.json();
+        throw new Error(error.message || "Transcription failed");
+      }
+
+      const { text, segments } = await transcriptionResponse.json();
+
+      console.log("Transcription completed:", { text, segments });
+
+      const shortCaptions: Array<{
+        text: string;
+        startTime: number;
+        duration: number;
+      }> = [];
+
+      let globalEndTime = 0; // Track the end time of the last caption globally
+
+      segments.forEach((segment: any) => {
+        const words = segment.text.trim().split(/\s+/);
+        const segmentDuration = segment.end - segment.start;
+        const wordsPerSecond = words.length / segmentDuration;
+
+        // Split into chunks of 2-4 words
+        const chunks: string[] = [];
+        for (let i = 0; i < words.length; i += 3) {
+          chunks.push(words.slice(i, i + 3).join(" "));
+        }
+
+        // Calculate timing for each chunk to place them sequentially
+        let chunkStartTime = segment.start;
+        chunks.forEach((chunk) => {
+          const chunkWords = chunk.split(/\s+/).length;
+          const chunkDuration = Math.max(0.8, chunkWords / wordsPerSecond); // Minimum 0.8s per chunk
+
+          let adjustedStartTime = chunkStartTime;
+
+          // Prevent overlapping: if this caption would start before the last one ends,
+          // start it right after the last one ends
+          if (adjustedStartTime < globalEndTime) {
+            adjustedStartTime = globalEndTime;
+          }
+
+          shortCaptions.push({
+            text: chunk,
+            startTime: adjustedStartTime,
+            duration: chunkDuration,
+          });
+
+          // Update global end time
+          globalEndTime = adjustedStartTime + chunkDuration;
+
+          // Next chunk starts when this one ends (for within-segment timing)
+          chunkStartTime += chunkDuration;
+        });
+      });
+
+      // Create a single track for all captions
+      const captionTrackId = insertTrackAt("text", 0);
+
+      // Add all caption elements to the same track
+      shortCaptions.forEach((caption, index) => {
+        addElementToTrack(captionTrackId, {
+          type: "text",
+          name: `Caption ${index + 1}`,
+          content: caption.text,
+          duration: caption.duration,
+          startTime: caption.startTime,
+          trimStart: 0,
+          trimEnd: 0,
+          fontSize: 65,
+          fontFamily: "Arial",
+          color: "#ffffff",
+          textAlign: "center",
+          fontWeight: "bold",
+          fontStyle: "normal",
+          textDecoration: "none",
+          x: 0,
+          y: 0,
+          rotation: 0,
+          opacity: 1,
+        } as TextElement);
+      });
+
+      console.log(
+        `✅ ${shortCaptions.length} short-form caption chunks added to timeline!`
+      );
+    } catch (error) {
+      console.error("Transcription failed:", error);
+      setError(
+        error instanceof Error ? error.message : "An unexpected error occurred"
+      );
+    } finally {
+      setIsProcessing(false);
+      setProcessingStep("");
+    }
+  };
+
+  return (
+    <BaseView ref={containerRef} className="flex flex-col justify-between">
+      <PropertyGroup title="Language">
+        <LanguageSelect
+          selectedCountry={selectedCountry}
+          onSelect={setSelectedCountry}
+          containerRef={containerRef}
+          languages={languages}
+        />
+      </PropertyGroup>
+
+      <div className="flex flex-col gap-4">
+        {error && (
+          <div className="p-3 bg-destructive/10 border border-destructive/20 rounded-md">
+            <p className="text-sm text-destructive">{error}</p>
+          </div>
+        )}
+
+        <Button
+          className="w-full"
+          onClick={() => {
+            if (hasAcceptedPrivacy) {
+              handleGenerateTranscript();
+            } else {
+              setShowPrivacyDialog(true);
+            }
+          }}
+          disabled={isProcessing}
+        >
+          {isProcessing && <Loader2 className="mr-1 h-4 w-4 animate-spin" />}
+          {isProcessing ? processingStep : "Generate transcript"}
+        </Button>
+
+        <Dialog open={showPrivacyDialog} onOpenChange={setShowPrivacyDialog}>
+          <DialogContent className="sm:max-w-md">
+            <DialogHeader>
+              <DialogTitle className="flex items-center gap-2">
+                <Shield className="h-5 w-5" />
+                Audio Processing Notice
+              </DialogTitle>
+              <DialogDescription className="space-y-3">
+                <p>
+                  To generate captions, we need to process your timeline audio
+                  using speech-to-text technology.
+                </p>
+
+                <div className="space-y-2 pt-2">
+                  <div className="flex items-start gap-2">
+                    <Shield className="h-4 w-4 flex-shrink-0" />
+                    <span className="text-sm">
+                      Zero-knowledge encryption - we cannot decrypt your files
+                      even if we wanted to
+                    </span>
+                  </div>
+
+                  <div className="flex items-start gap-2">
+                    <Shield className="h-4 w-4 flex-shrink-0" />
+                    <span className="text-sm">
+                      Encryption keys generated randomly in your browser, never
+                      stored anywhere
+                    </span>
+                  </div>
+
+                  <div className="flex items-start gap-2">
+                    <Upload className="h-4 w-4 flex-shrink-0" />
+                    <span className="text-sm">
+                      Audio encrypted before upload - raw audio never leaves
+                      your device
+                    </span>
+                  </div>
+
+                  <div className="flex items-start gap-2">
+                    <Trash2 className="h-4 w-4 flex-shrink-0" />
+                    <span className="text-sm">
+                      Everything permanently deleted within seconds after
+                      transcription
+                    </span>
+                  </div>
+                </div>
+
+                <p className="text-xs text-muted-foreground">
+                  <strong>True zero-knowledge privacy:</strong> Encryption keys
+                  are generated randomly in your browser and never stored
+                  anywhere. It's cryptographically impossible for us, our cloud
+                  providers, or anyone else to decrypt your audio files.
+                </p>
+              </DialogDescription>
+            </DialogHeader>
+            <DialogFooter className="gap-2">
+              <Button
+                variant="outline"
+                onClick={() => setShowPrivacyDialog(false)}
+                disabled={isProcessing}
+              >
+                Cancel
+              </Button>
+              <Button
+                onClick={() => {
+                  localStorage.setItem(PRIVACY_DIALOG_KEY, "true");
+                  setHasAcceptedPrivacy(true);
+                  setShowPrivacyDialog(false);
+                  handleGenerateTranscript();
+                }}
+                disabled={isProcessing}
+              >
+                Continue & Generate Captions
+              </Button>
+            </DialogFooter>
+          </DialogContent>
+        </Dialog>
+      </div>
+    </BaseView>
+  );
+}
-- 
cgit v1.3.1