diff --git a/engine/core/Engine.ts b/engine/core/Engine.ts
index eed9cec..89151f8 100644
--- a/engine/core/Engine.ts
+++ b/engine/core/Engine.ts
@@ -107,10 +107,14 @@ export class Engine {
     this.emit('sceneChange', scene)
     this.checkHotspotTime(scene, 0)
 
-    const preloadUrls = this.sceneManager.getCandidateUrls(
+    const candidateIds = this.sceneManager.getCandidateSceneIds(
       scene,
       (conds) => conds ? this.stateManager.evaluate(conds) : true
     )
+    const preloadUrls = candidateIds
+      .map(id => this.sceneManager.getScene(id))
+       .filter((s) => !!s && s.type !== 'image')
+       .map(s => this.videoManager.resolveVideoUrl(s!, this.videoManager.streamingQuality))
 
     this.videoManager.onEnd(() => {
       if (!this.qteTriggered || this.qteResolved) {
diff --git a/engine/core/SceneManager.ts b/engine/core/SceneManager.ts
index 1e20cd4..a982443 100644
--- a/engine/core/SceneManager.ts
+++ b/engine/core/SceneManager.ts
@@ -57,9 +57,7 @@ export class SceneManager {
     return targets
   }
 
-  getCandidateUrls(scene: SceneNode, evaluateCondition: (conds?: Condition[]) => boolean): string[] {
+  getCandidateSceneIds(scene: SceneNode, evaluateCondition: (conds?: Condition[]) => boolean): string[] {
     return this.getCandidateTargetIds(scene, evaluateCondition)
-      .map(id => this.scenes[id]?.videoUrl)
-      .filter((url): url is string => !!url)
   }
 }