feat: add generation time display

- Added response time tracking in infer_script.py
- Display actual generation time in UI with formatted output
- Shows time in seconds or minutes based on duration

app.py

@@ -251,7 +251,7 @@ def generate_video(input_img, should_crop_face, expand_x, expand_y, offset_x, of
 
     try:
         print("📞 Calling run_inference...")
-        status, oo_video_path, all_video_path = run_inference(
+        status, oo_video_path, all_video_path, elapsed_time = run_inference(
             config_path=INFER_CONFIG_PATH,
             model_path=MODEL_PATH,
             input_path=input_img,
@@ -274,6 +274,7 @@ def generate_video(input_img, should_crop_face, expand_x, expand_y, offset_x, of
             interpolation_factor=intropolate_factor
         )
         print(f"✅ Inference completed! Status: {status}")
+        print(f"⏱️ Total time: {elapsed_time:.2f} seconds")
     except Exception as e:
         print(f"❌ Error in run_inference: {str(e)}")
         import traceback
@@ -296,7 +297,15 @@ def generate_video(input_img, should_crop_face, expand_x, expand_y, offset_x, of
     else:
         print(f"Directory with frames not found: {frames_dir}")
 
-    return status, oo_video_path, all_video_path, frames_archive
+    # Format timing display
+    if elapsed_time < 60:
+        time_display = f"⏱️ {elapsed_time:.2f} seconds"
+    else:
+        minutes = int(elapsed_time // 60)
+        seconds = elapsed_time % 60
+        time_display = f"⏱️ {minutes} min {seconds:.2f} sec"
+
+    return status, time_display, oo_video_path, all_video_path, frames_archive
 
 with gr.Blocks() as demo:
     gr.Markdown("<h1 style='text-align: center;'>FollowYourEmoji</h1>")
@@ -378,6 +387,7 @@ with gr.Blocks() as demo:
         with gr.Column(scale=1):
             gr.Markdown("### 🎨 Generated Results")
             result_status = gr.Label(value="Ready to generate")
+            result_time = gr.Label(value="Response time will appear here", label="⏱️ Generation Time")
             result_video = gr.Video(label="Result Video (Main)", interactive=False, height=400)
             result_video_2 = gr.Video(label="Result Video (Full)", interactive=False, height=400)
             frames_output = gr.File(label="Download Frames Archive")
@@ -460,7 +470,7 @@ with gr.Blocks() as demo:
         inputs=[input_img, crop_face_checkbox, expand_x, expand_y, offset_x, offset_y, input_video_type, input_video, input_npy_select, input_npy, input_video_frames,
             settings_steps, settings_cfg_scale, settings_seed, resolution_w, resolution_h,
             model_step, custom_output_path, use_custom_fps, output_fps, callback_steps, context_frames, context_stride, context_overlap, context_batch_size, anomaly_action,intropolate_factor],
-        outputs=[result_status, result_video, result_video_2, frames_output],
+        outputs=[result_status, result_time, result_video, result_video_2, frames_output],
         show_progress="full"
     )
     

infer_script.py

@@ -172,6 +172,8 @@ def visualize(dataloader, pipeline, generator, W, H, video_length, num_inference
 def infer(config_path, model_path, input_path, lmk_path, output_path, model_step, seed,
           resolution_w, resolution_h, video_length, num_inference_steps, guidance_scale, output_fps, show_stats,
           anomaly_action, callback_steps, context_frames, context_stride, context_overlap, context_batch_size,interpolation_factor):
+    import time
+    start_time = time.time()
 
     config = OmegaConf.load(config_path)
     config.init_checkpoint = model_path
@@ -274,7 +276,18 @@ def infer(config_path, model_path, input_path, lmk_path, output_path, model_step
     torch.cuda.empty_cache()
     gc.collect()
 
-    return "Inference completed successfully", oo_video_path, all_video_path
+    end_time = time.time()
+    elapsed_time = end_time - start_time
+    
+    # Format time nicely
+    if elapsed_time < 60:
+        time_str = f"{elapsed_time:.2f} seconds"
+    else:
+        minutes = int(elapsed_time // 60)
+        seconds = elapsed_time % 60
+        time_str = f"{minutes} min {seconds:.2f} sec"
+
+    return f"Inference completed successfully in {time_str}", oo_video_path, all_video_path, elapsed_time
 
 def run_inference(config_path, model_path, input_path, lmk_path, output_path, model_step, seed,
                   resolution_w, resolution_h, video_length, num_inference_steps=30, guidance_scale=3.5, output_fps=30,
@@ -285,9 +298,10 @@ def run_inference(config_path, model_path, input_path, lmk_path, output_path, mo
         torch.cuda.empty_cache()
         gc.collect()
 
-        return infer(config_path, model_path, input_path, lmk_path, output_path, model_step, seed,
+        status, oo_video_path, all_video_path, elapsed_time = infer(config_path, model_path, input_path, lmk_path, output_path, model_step, seed,
                      resolution_w, resolution_h, video_length, num_inference_steps, guidance_scale, output_fps,
                      show_stats, anomaly_action, callback_steps, context_frames, context_stride, context_overlap, context_batch_size,interpolation_factor)
+        return status, oo_video_path, all_video_path, elapsed_time
     finally:
         torch.cuda.empty_cache()
         gc.collect()
@@ -320,7 +334,7 @@ if __name__ == "__main__":
 
     args = parser.parse_args()
 
-    status, oo_path, all_path = run_inference(
+    status, oo_path, all_path, elapsed_time = run_inference(
         args.config, args.model, args.input, args.lmk, args.output, args.step, args.seed,
         args.width, args.height, args.length, args.steps, args.guidance, args.fps,
         args.show_stats, args.anomaly_action, args.callback_steps, args.context_frames,
@@ -330,3 +344,4 @@ if __name__ == "__main__":
     print(status)
     print(f"Output video (only output): {oo_path}")
     print(f"Output video (all frames): {all_path}")
+    print(f"Total generation time: {elapsed_time:.2f} seconds")