update app

app.py

@@ -138,7 +138,7 @@ css = """
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 print(f"Using Main Device: {device}")
 
-QWEN_VL_MODEL_ID = "Qwen/Qwen2.5-VL-7B-Instruct"
+QWEN_VL_MODEL_ID = "Qwen/Qwen2.5-VL-3B-Instruct"
 print(f"Loading OCR Model: {QWEN_VL_MODEL_ID}...")
 
 qwen_processor = AutoProcessor.from_pretrained(QWEN_VL_MODEL_ID, trust_remote_code=True)
@@ -335,7 +335,7 @@ example_image = Image.open(requests.get(url, stream=True).raw).convert("RGB")
 
 with gr.Blocks() as demo:
     gr.Markdown("# **Vision-to-VibeVoice-en**", elem_id="main-title")
-    gr.Markdown("Perform vision-to-audio inference with [Qwen2.5VL](https://huggingface.co/Qwen/Qwen2.5-VL-7B-Instruct) + [VibeVoice-Realtime-0.5B](https://huggingface.co/microsoft/VibeVoice-Realtime-0.5B).")
+    gr.Markdown("Perform vision-to-audio inference with [Qwen2.5VL](https://huggingface.co/Qwen/Qwen2.5-VL-3B-Instruct) + [VibeVoice-Realtime-0.5B](https://huggingface.co/microsoft/VibeVoice-Realtime-0.5B).")
     with gr.Row():
         with gr.Column(scale=1):
             gr.Markdown("### 1. Vision Input")
@@ -355,7 +355,7 @@ with gr.Blocks() as demo:
             cfg_slider = gr.Slider(minimum=1.0, maximum=3.0, value=1.5, step=0.1, label="CFG Scale (Speech Fidelity)")
                 
             with gr.Accordion("Advanced Options", open=False):
-                max_new_tokens = gr.Slider(label="Max Tokens", minimum=128, maximum=4096, step=128, value=2048)
+                max_new_tokens = gr.Slider(label="Max Tokens", minimum=128, maximum=4096, step=128, value=1024)
                 temperature = gr.Slider(label="Temperature", minimum=0.1, maximum=2.0, step=0.1, value=0.1)
 
             submit_btn = gr.Button("Generate Speech", variant="primary", size="lg")