testing something

app.py

@@ -225,23 +225,22 @@ pipeline.to(dtype=torch.float16).to(gpu)
 def video_generation_handler_streaming(prompt, seed=42, fps=15, width=400, height=224, duration=5, buffering=2):
     """
     Generator function that yields .ts video chunks using PyAV for streaming.
-    Now optimized for block-based processing.
+    Now optimized for block-based processing with smart buffering.
     """
     if seed == -1: 
         seed = random.randint(0, 2**32 - 1)
     
     print(f"🎬 Starting PyAV streaming: '{prompt}', seed: {seed}, duration: {duration}s, buffering: {buffering}s")
     
-    # Buffering delay
+    # Show initial buffering status but don't wait - start generating immediately
     if buffering > 0:
         buffering_status_html = (
             f"<div style='padding: 10px; border: 1px solid #ffc107; background: #fff3cd; border-radius: 8px; font-family: sans-serif;'>"
             f"  <p style='margin: 0 0 8px 0; font-size: 16px; font-weight: bold;'>⏳ Buffering...</p>"
-            f"  <p style='margin: 0; color: #856404; font-size: 14px;'>Waiting {buffering} seconds before starting stream</p>"
+            f"  <p style='margin: 0; color: #856404; font-size: 14px;'>Generating content, will stream when {buffering} seconds of video is ready</p>"
             f"</div>"
         )
         yield None, buffering_status_html
-        time.sleep(buffering)
     
     # Setup
     conditional_dict = text_encoder(text_prompts=[prompt])
@@ -272,6 +271,12 @@ def video_generation_handler_streaming(prompt, seed=42, fps=15, width=400, heigh
     # Ensure temp directory exists
     os.makedirs("gradio_tmp", exist_ok=True)
     
+    # Buffer management - collect chunks before streaming
+    buffer_chunks = []
+    buffer_duration = 0.0
+    frames_per_second = fps
+    streaming_started = False
+    
     # Generation loop
     for idx, current_num_frames in enumerate(all_num_frames):
         print(f"📦 Processing block {idx+1}/{num_blocks}")
@@ -359,7 +364,7 @@ def video_generation_handler_streaming(prompt, seed=42, fps=15, width=400, heigh
             # Yield None for video but update status (frame-by-frame tracking)
             yield None, frame_status_html
 
-        # Encode entire block as one chunk immediately
+        # Encode entire block as one chunk
         if all_frames_from_block:
             print(f"📹 Encoding block {idx} with {len(all_frames_from_block)} frames")
             
@@ -370,11 +375,45 @@ def video_generation_handler_streaming(prompt, seed=42, fps=15, width=400, heigh
                 
                 frames_to_ts_file(all_frames_from_block, ts_path, fps)
                 
-                # Calculate final progress for this block
-                total_progress = (idx + 1) / num_blocks * 100
+                # Calculate duration of this chunk
+                chunk_duration = len(all_frames_from_block) / frames_per_second
+                
+                # Add to buffer
+                buffer_chunks.append(ts_path)
+                buffer_duration += chunk_duration
                 
-                # Yield the actual video chunk
-                yield ts_path, gr.update()
+                # Check if we have enough buffered content to start streaming
+                if not streaming_started and buffer_duration >= buffering:
+                    print(f"🚀 Buffer filled ({buffer_duration:.2f}s >= {buffering}s), starting stream!")
+                    streaming_started = True
+                    
+                    # Stream all buffered chunks
+                    for buffered_chunk in buffer_chunks:
+                        yield buffered_chunk, gr.update()
+                    
+                    # Clear buffer since we've streamed it
+                    buffer_chunks.clear()
+                    buffer_duration = 0.0
+                    
+                elif streaming_started:
+                    # Stream immediately if we're already streaming
+                    yield ts_path, gr.update()
+                elif buffering == 0:
+                    # No buffering requested, stream immediately
+                    yield ts_path, gr.update()
+                else:
+                    # Still buffering, show progress
+                    buffer_progress = (buffer_duration / buffering) * 100
+                    buffering_progress_html = (
+                        f"<div style='padding: 10px; border: 1px solid #ffc107; background: #fff3cd; border-radius: 8px; font-family: sans-serif;'>"
+                        f"  <p style='margin: 0 0 8px 0; font-size: 16px; font-weight: bold;'>⏳ Buffering... ({buffer_duration:.1f}s/{buffering}s)</p>"
+                        f"  <div style='background: #e9ecef; border-radius: 4px; width: 100%; overflow: hidden;'>"
+                        f"    <div style='width: {buffer_progress:.1f}%; height: 20px; background-color: #ffc107; transition: width 0.2s;'></div>"
+                        f"  </div>"
+                        f"  <p style='margin: 4px 0 0 0; color: #856404; font-size: 14px;'>Generating content for smooth playback...</p>"
+                        f"</div>"
+                    )
+                    yield None, buffering_progress_html
                 
             except Exception as e:
                 print(f"⚠️ Error encoding block {idx}: {e}")
@@ -383,6 +422,12 @@ def video_generation_handler_streaming(prompt, seed=42, fps=15, width=400, heigh
                     
         current_start_frame += current_num_frames
     
+    # Stream any remaining buffered content
+    if buffer_chunks:
+        print(f"🎬 Streaming remaining {len(buffer_chunks)} buffered chunks")
+        for buffered_chunk in buffer_chunks:
+            yield buffered_chunk, gr.update()
+    
     # Final completion status
     final_status_html = (
         f"<div style='padding: 16px; border: 1px solid #198754; background: linear-gradient(135deg, #d1e7dd, #f8f9fa); border-radius: 8px; box-shadow: 0 2px 4px rgba(0,0,0,0.1);'>"