Update app.py

app.py

@@ -26,12 +26,12 @@ def load_pipelines():
             device=0 if torch.cuda.is_available() else -1
         )
 
-        # # 3. Soundtrack generation (MusicGen)
-        # # music_pipe = pipeline(
-        # #     "text-to-audio", 
-        # #     model="facebook/musicgen-small",
-        # #     device_map="auto"
-        # # )
+        # 3. Soundtrack generation (MusicGen)
+        music_pipe = pipeline(
+            "text-to-audio", 
+            model="facebook/musicgen-small",
+            device_map="auto"
+        )
 
         # 4. Storyboard image generation (Stable Diffusion)
         image_pipe = StableDiffusionPipeline.from_pretrained(
@@ -236,4 +236,44 @@ if user_input and style_or_opening:
     #     st.error("生成过程中发生错误")
     #     st.code(f"错误详情：{str(e)}")
     #     st.button("点击查看技术详情", 
-    #              help="将以下信息提供给技术人员：\n" + str(e))
+    #              help="将以下信息提供给技术人员：\n" + str(e))
+    st.subheader("🎧 Background Music")
+    for i, scene in enumerate(parsed["scenes"]):
+        action = scene["action"]
+        with st.spinner(f"🎵 Generating soundtrack for scene {i + 1}..."):
+            try:
+                audio = music_pipe(
+                    action,
+                    max_new_tokens=256
+                )
+                audio_buffer = BytesIO()
+                audio["audio"].export(audio_buffer, format="wav")
+                st.audio(audio_buffer, format="audio/wav", caption=f"Soundtrack for scene {i + 1}")
+            except Exception as e:
+                st.error(f"Failed to generate soundtrack for scene {i + 1}: {str(e)}")
+
+    st.subheader("🎨 Storyboard Images")
+    for i, scene in enumerate(parsed["scenes"]):
+        location = scene["location"]
+        if st.toggle(f"Generate preview for scene {i + 1} (takes 2 - 3 minutes)"):
+            with st.spinner(f"🖼️ Rendering image for scene {i + 1}..."):
+                try:
+                    # 精细化图片提示词
+                    image_prompt = f"""Movie still, {location},
+                        Ultra - high - definition 8K resolution,
+                        Movie - level lighting, shallow depth - of - field effect,
+                        Unreal Engine 5 rendering,
+                        Movie film graininess"""
+
+                    # 优化生成参数
+                    image = image_pipe(
+                        image_prompt,
+                        num_inference_steps=50,
+                        guidance_scale=9.5,
+                        width=1024,
+                        height=576  # 16:9 movie ratio
+                    ).images[0]
+
+                    st.image(image, caption=f"Preview for scene {i + 1}")
+                except Exception as e:
+                    st.error(f"Failed to generate image for scene {i + 1}: {str(e)}")