fix: pass image as video input for lipsync models (sync-v2-pro requires video_url)

Author: SecurityQQ

src/react/renderers/talking-head.ts

@@ -54,12 +54,16 @@ export async function renderTalkingHead(
   const characterImageData = await characterFile.arrayBuffer();
   const speechAudioData = await speechFile.arrayBuffer();
 
-  // Create a synthetic video element for the lipsync generation
+  // Create a synthetic video element for the lipsync generation.
+  // Lipsync models (sync-v2-pro, etc.) require `video_url`, not `image_url`,
+  // so we pass the character image as the `video` input. The fal provider will
+  // upload it and set `video_url` in the API request. Fal.ai accepts image
+  // files as the video input for lipsync — it treats them as single-frame video.
   const videoElement: VargElement<"video"> = {
     type: "video",
     props: {
       prompt: {
-        images: [characterImageData],
+        video: characterImageData,
         audio: speechAudioData,
       },
       model: lipsyncModel,

src/react/resolve.ts

@@ -808,10 +808,12 @@ export async function resolveTalkingHeadElement(
   const lipsyncModel = props.lipsyncModel ?? model;
   const generateVideo = getCachedGenerateVideo();
 
+  // Lipsync models require `video_url`, not `image_url`, so pass the
+  // character image as the `video` input (fal accepts images as video input).
   const { video } = await generateVideo({
     model: lipsyncModel as Parameters<typeof generateVideoRaw>[0]["model"],
     prompt: {
-      images: [characterBytes],
+      video: characterBytes,
       audio: speechBytes,
     },
     duration: 0, // duration determined by audio length