Update app.py

app.py

@@ -865,7 +865,7 @@ def _extract_audio_from_video(video_path: str) -> str | None:
         # Extract audio
         subprocess.run(
             ["ffmpeg", "-y", "-v", "error", "-i", video_path,
-             "-vn", "-ac", "1", "-ar", "48000", "-c:a", "pcm_s16le", out_path],
+             "-vn", "-ac", "2", "-ar", "48000", "-c:a", "pcm_s16le", out_path],
             check=True,
         )
         return out_path
@@ -884,6 +884,7 @@ def generate_video(
     conditioning_strength: float,
     video_preprocess: str,
     enhance_prompt: bool,
+    use_video_audio: bool,
     seed: int,
     randomize_seed: bool,
     height: int,
@@ -945,8 +946,8 @@ def generate_video(
                 # Raw mode — pass video as-is
                 video_conditioning = [(video_path, 1.0)]
 
-            # If no audio was provided, try to extract audio from the video
-            if input_audio is None:
+            # If no audio was provided, optionally extract audio from the video
+            if input_audio is None and use_video_audio:
                 extracted_audio = _extract_audio_from_video(video_path)
                 if extracted_audio is not None:
                     input_audio = extracted_audio
@@ -1027,11 +1028,6 @@ with gr.Blocks(title="LTX-2.3 Unified: V2V + I2V + A2V") as demo:
                 label="🔊 Input Audio (A2V — lipsync / BGM)",
                 type="filepath",
             )
-            gr.Markdown(
-                "*When a video is uploaded: its first frame auto-becomes the image input "
-                "(if none provided), and its audio track auto-becomes the audio input "
-                "(if none provided).*"
-            )
 
             prompt = gr.Textbox(
                 label="Prompt",
@@ -1055,6 +1051,11 @@ with gr.Blocks(title="LTX-2.3 Unified: V2V + I2V + A2V") as demo:
             with gr.Row():
                 enhance_prompt = gr.Checkbox(label="Enhance Prompt", value=True)
                 high_res = gr.Checkbox(label="High Resolution", value=True)
+                use_video_audio = gr.Checkbox(
+                    label="Use Audio from Video",
+                    value=True,
+                    info="Extract and use the audio track from the reference video",
+                )
 
             generate_btn = gr.Button("Generate Video", variant="primary", size="lg")
 
@@ -1091,7 +1092,7 @@ with gr.Blocks(title="LTX-2.3 Unified: V2V + I2V + A2V") as demo:
         inputs=[
             input_image, input_video, input_audio, prompt, duration,
             conditioning_strength, video_preprocess, enhance_prompt,
-            seed, randomize_seed, height, width,
+            use_video_audio, seed, randomize_seed, height, width,
         ],
         outputs=[output_video, seed],
     )