vargai 0.4.0-alpha63 → 0.4.0-alpha65

package/package.json

@@ -70,7 +70,7 @@
     "zod": "^4.2.1"
   },
   "sideEffects": false,
-  "version": "0.4.0-alpha63",
+  "version": "0.4.0-alpha65",
   "exports": {
     ".": "./src/index.ts",
     "./ai": "./src/ai-sdk/index.ts",

package/src/ai-sdk/file.ts

@@ -268,6 +268,8 @@ export class File {
       "video/mp4": ".mp4",
       "video/webm": ".webm",
       "video/quicktime": ".mov",
+      "text/x-ssa": ".ass",
+      "application/x-subrip": ".srt",
     };
     return extMap[this._mediaType] ?? "";
   }
@@ -290,6 +292,9 @@ function inferMediaType(path: string): string {
     mp4: "video/mp4",
     webm: "video/webm",
     mov: "video/quicktime",
+    ass: "text/x-ssa",
+    ssa: "text/x-ssa",
+    srt: "application/x-subrip",
   };
   return mimeTypes[ext ?? ""] ?? "application/octet-stream";
 }

package/src/react/renderers/render.ts

@@ -170,6 +170,30 @@ export async function renderRoot(
   const audioTracks: AudioTrack[] = [];
   let captionsResult: Awaited<ReturnType<typeof renderCaptions>> | undefined;
 
+  // Hoist <Captions> out of <Clip> elements — the AI often places them inside
+  // clips, but captions must be processed at the <Render> level to work.
+  const hoistedCaptions: VargElement<"captions">[] = [];
+  for (const child of element.children) {
+    if (!child || typeof child !== "object" || !("type" in child)) continue;
+    const childElement = child as VargElement;
+    if (childElement.type === "clip" && childElement.children) {
+      const kept: typeof childElement.children = [];
+      for (const clipChild of childElement.children) {
+        if (
+          clipChild &&
+          typeof clipChild === "object" &&
+          "type" in clipChild &&
+          (clipChild as VargElement).type === "captions"
+        ) {
+          hoistedCaptions.push(clipChild as VargElement<"captions">);
+        } else {
+          kept.push(clipChild);
+        }
+      }
+      childElement.children = kept;
+    }
+  }
+
   for (const child of element.children) {
     if (!child || typeof child !== "object" || !("type" in child)) continue;
 
@@ -206,6 +230,18 @@ export async function renderRoot(
     }
   }
 
+  // Process any <Captions> that were hoisted from inside <Clip> elements
+  if (!captionsResult && hoistedCaptions.length > 0) {
+    const captionsElement = hoistedCaptions[0]!;
+    captionsResult = await renderCaptions(captionsElement, ctx);
+    if (captionsResult.audioPath) {
+      audioTracks.push({
+        path: captionsResult.audioPath,
+        mixVolume: 1,
+      });
+    }
+  }
+
   const renderedOverlays: RenderedOverlay[] = [];
   for (const overlay of overlayElements) {
     const overlayProps = overlay.props as OverlayProps;