update

modal_video_processing.py

@@ -41,7 +41,14 @@ def process_quote_video(
     - Overlays `quote_text` using a chosen `text_style`.
     - If `audio_b64` is provided, decodes it and:
         * sets it as the audio track
-        * loops the video so its duration is at least the audio duration (+ small buffer).
+        * makes video duration roughly match audio (with min/max bounds).
+
+    Duration rules:
+    - With audio:
+        target = audio_duration + 0.5s
+        MIN = 7s, MAX = 20s
+    - Without audio:
+        target = min(original_video_duration, 15s)
 
     Returns:
         Raw bytes of the final MP4 video.
@@ -77,13 +84,17 @@ def process_quote_video(
     # 2. Load video
     # ---------------------------
     video = VideoFileClip(temp_video.name)
+    orig_duration = video.duration
 
     # ---------------------------
-    # 3. Optional audio handling
+    # 3. Duration logic + optional audio
     # ---------------------------
     audio_clip = None
     temp_audio_path = None
 
+    # Default target when no audio
+    target_duration = orig_duration
+
     if audio_b64:
         try:
             temp_audio = tempfile.NamedTemporaryFile(delete=False, suffix=".mp3")
@@ -97,20 +108,36 @@ def process_quote_video(
             audio_clip = AudioFileClip(temp_audio_path)
             audio_duration = audio_clip.duration
 
-            # Small buffer so we don't cut off the last word
-            target_duration = audio_duration + 0.5
+            # Proportional rules with audio
+            MIN_DUR = 7.0
+            MAX_DUR = 20.0
+            target_duration = audio_duration + 0.5  # small buffer
+            if target_duration < MIN_DUR:
+                target_duration = MIN_DUR
+            if target_duration > MAX_DUR:
+                target_duration = MAX_DUR
 
-            # If audio is longer than the clip, loop the video to at least cover it
+            # Adjust video to target_duration
             if target_duration > video.duration:
                 video = vfx_loop(video, duration=target_duration)
+            elif target_duration < video.duration:
+                video = video.subclip(0, target_duration)
 
-            # IMPORTANT: we do NOT crop video when audio is shorter;
-            # it's fine if video runs a bit longer than the narration.
         except Exception as e:
             print(f"⚠️ Audio handling error: {e}")
             audio_clip = None
+            # Fall back to no-audio behavior below
+
+    if audio_clip is None:
+        # No audio path: clamp to reasonable length
+        MAX_NO_AUDIO = 15.0
+        if orig_duration > MAX_NO_AUDIO:
+            target_duration = MAX_NO_AUDIO
+            video = video.subclip(0, target_duration)
+        else:
+            target_duration = orig_duration
 
-    # If no audio, keep original video duration
+    # At this point, video.duration ≈ target_duration
     w, h = video.size
 
     # ---------------------------
@@ -193,11 +220,10 @@ def process_quote_video(
     # ---------------------------
     final_video = CompositeVideoClip([video, text_clip])
 
-    # If we have audio, attach it (no extra duration forcing)
+    # Attach audio if available (no extra duration forcing)
     if audio_clip is not None:
         try:
             final_video = final_video.set_audio(audio_clip)
-            # Do NOT call set_duration() again: video is already long enough.
         except Exception as e:
             print(f"⚠️ Could not attach audio: {e}")
 
@@ -254,7 +280,8 @@ def process_quote_video(
 
     total_time = time.time() - start_time
     print(
-        f"🎉 Total: {total_time:.1f}s, Size: {len(video_bytes) / 1024 / 1024:.2f}MB, text_style={text_style}"
+        f"🎉 Total: {total_time:.1f}s, Size: {len(video_bytes) / 1024 / 1024:.2f}MB, "
+        f"text_style={text_style}, target_duration≈{target_duration:.1f}s"
     )
 
     return video_bytes