Update app.py

app.py

@@ -74,6 +74,16 @@ def load_sample_shot(camera_shot):
 def use_custom_cond():
     return "custom"
 
+def get_short_caption(image_in):
+    client = Client("https://vikhyatk-moondream1.hf.space/")
+    result = client.predict(
+		image_in,	# filepath  in 'image' Image component
+		"Describe what is happening in one sentence",	# str  in 'Question' Textbox component
+		api_name="/answer_question"
+    )
+    print(result)
+    return result
+
 def infer(image_in, camera_shot, conditional_pose, prompt, style, chosen_model):
     
     if camera_shot == "custom":
@@ -84,9 +94,11 @@ def infer(image_in, camera_shot, conditional_pose, prompt, style, chosen_model):
     
     
     iid_img = get_instantID(image_in, conditional_pose, prompt, style)
+
+    short_cap = get_short_caption(iid_img)
     
     if chosen_model == "i2vgen-xl" :
-        video_res = get_video_i2vgen(iid_img, prompt)
+        video_res = get_video_i2vgen(iid_img, short_cap)
     elif chosen_model == "stable-video" :
         video_res = get_video_svd(image_in)
     
@@ -128,8 +140,8 @@ with gr.Blocks(css=css) as demo:
                     )
                     style = gr.Dropdown(label="Style template", info="InstantID legacy templates", choices=STYLE_NAMES, value=DEFAULT_STYLE_NAME)
 
-                condition_shot = gr.Image(type="filepath", label="Custom conditional shot (Optional)")
-                prompt = gr.Textbox(label="Short Prompt (keeping it short is better")
+                condition_shot = gr.Image(type="filepath", label="Custom conditional shot (Important) [1280*720 recommended]")
+                prompt = gr.Textbox(label="Short Prompt (keeping it short is better)")
                 chosen_model = gr.Radio(label="Choose a model", choices=["i2vgen-xl", "stable-video"], value="i2vgen-xl", interactive=False, visible=False)
             
         with gr.Column():