Integration of cropdetec-logic into anime_audio_encoder and tv_audio_encoder. Rewrite of readmes

tv_audio_encoder.py

@@ -105,7 +105,7 @@ def convert_audio_track(index, ch, lang, audio_temp_dir, source_file, should_dow
     ])
     return final_opus
 
-def convert_video(source_file_base, source_file_full):
+def convert_video(source_file_base, source_file_full, autocrop_filter=None):
     print("  --- Starting Video Processing ---")
     # source_file_base is the full stem from the original file,
     # e.g., "cheers.s01e04.der.lueckenbuesser.german.dl.fs.1080p.web.h264-cnhd"
@@ -129,7 +129,10 @@ def convert_video(source_file_base, source_file_full):
     ffmpeg_args = [
         "ffmpeg", "-hide_banner", "-v", "quiet", "-stats", "-y", "-i", source_file_full,
         "-map", "0:v:0", "-map_metadata", "-1", "-map_chapters", "-1", "-an", "-sn", "-dn",
-    ] + video_codec_args + [str(ut_video_file)]
+    ]
+    if autocrop_filter:
+        ffmpeg_args += ["-vf", autocrop_filter]
+    ffmpeg_args += video_codec_args + [str(ut_video_file)]
     run_cmd(ffmpeg_args)
 
     print("    - Starting video encode with AlabamaEncoder (this will take a long time)...")
@@ -159,58 +162,263 @@ def convert_video(source_file_base, source_file_full):
     print("  --- Finished Video Processing ---")
     return ut_video_file, encoded_video_file
 
-def main(no_downmix=False):
+# --- CROPDETECT LOGIC FROM cropdetect.py ---
+import multiprocessing as _multiprocessing_cropdetect
+from collections import Counter as _Counter_cropdetect
+
+KNOWN_ASPECT_RATIOS = [
+    {"name": "HDTV (16:9)", "ratio": 16/9},
+    {"name": "Widescreen (Scope)", "ratio": 2.39},
+    {"name": "Widescreen (Flat)", "ratio": 1.85},
+    {"name": "IMAX Digital (1.90:1)", "ratio": 1.90},
+    {"name": "Fullscreen (4:3)", "ratio": 4/3},
+    {"name": "IMAX 70mm (1.43:1)", "ratio": 1.43},
+]
+
+def _check_prerequisites_cropdetect():
+    for tool in ['ffmpeg', 'ffprobe']:
+        if not shutil.which(tool):
+            print(f"Error: '{tool}' command not found. Is it installed and in your PATH?")
+            return False
+    return True
+
+def _analyze_segment_cropdetect(task_args):
+    seek_time, input_file, width, height = task_args
+    ffmpeg_args = [
+        'ffmpeg', '-hide_banner',
+        '-ss', str(seek_time),
+        '-i', input_file, '-t', '1', '-vf', 'cropdetect',
+        '-f', 'null', '-'
+    ]
+    result = subprocess.run(ffmpeg_args, capture_output=True, text=True, encoding='utf-8')
+    if result.returncode != 0:
+        return []
+    import re
+    crop_detections = re.findall(r'crop=(\d+):(\d+):(\d+):(\d+)', result.stderr)
+    significant_crops = []
+    for w_str, h_str, x_str, y_str in crop_detections:
+        w, h, x, y = map(int, [w_str, h_str, x_str, y_str])
+        significant_crops.append((f"crop={w}:{h}:{x}:{y}", seek_time))
+    return significant_crops
+
+def _snap_to_known_ar_cropdetect(w, h, x, y, video_w, video_h, tolerance=0.03):
+    if h == 0: return f"crop={w}:{h}:{x}:{y}", None
+    detected_ratio = w / h
+    best_match = None
+    smallest_diff = float('inf')
+    for ar in KNOWN_ASPECT_RATIOS:
+        diff = abs(detected_ratio - ar['ratio'])
+        if diff < smallest_diff:
+            smallest_diff = diff
+            best_match = ar
+    if not best_match or (smallest_diff / best_match['ratio']) >= tolerance:
+        return f"crop={w}:{h}:{x}:{y}", None
+    if abs(w - video_w) < 16:
+        new_h = round(video_w / best_match['ratio'])
+        if new_h % 8 != 0:
+            new_h = new_h + (8 - (new_h % 8))
+        new_y = round((video_h - new_h) / 2)
+        if new_y % 2 != 0:
+            new_y -= 1
+        return f"crop={video_w}:{new_h}:0:{new_y}", best_match['name']
+    if abs(h - video_h) < 16:
+        new_w = round(video_h * best_match['ratio'])
+        if new_w % 8 != 0:
+            new_w = new_w + (8 - (new_w % 8))
+        new_x = round((video_w - new_w) / 2)
+        if new_x % 2 != 0:
+            new_x -= 1
+        return f"crop={new_w}:{video_h}:{new_x}:0", best_match['name']
+    return f"crop={w}:{h}:{x}:{y}", None
+
+def _cluster_crop_values_cropdetect(crop_counts, tolerance=8):
+    clusters = []
+    temp_counts = crop_counts.copy()
+    while temp_counts:
+        center_str, _ = temp_counts.most_common(1)[0]
+        try:
+            _, values = center_str.split('=');
+            cw, ch, cx, cy = map(int, values.split(':'))
+        except (ValueError, IndexError):
+            del temp_counts[center_str]
+            continue
+        cluster_total_count = 0
+        crops_to_remove = []
+        for crop_str, count in temp_counts.items():
+            try:
+                _, values = crop_str.split('=');
+                w, h, x, y = map(int, values.split(':'))
+                if abs(x - cx) <= tolerance and abs(y - cy) <= tolerance:
+                    cluster_total_count += count
+                    crops_to_remove.append(crop_str)
+            except (ValueError, IndexError):
+                continue
+        if cluster_total_count > 0:
+            clusters.append({'center': center_str, 'count': cluster_total_count})
+        for crop_str in crops_to_remove:
+            del temp_counts[crop_str]
+    clusters.sort(key=lambda c: c['count'], reverse=True)
+    return clusters
+
+def _parse_crop_string_cropdetect(crop_str):
+    try:
+        _, values = crop_str.split('=');
+        w, h, x, y = map(int, values.split(':'))
+        return {'w': w, 'h': h, 'x': x, 'y': y}
+    except (ValueError, IndexError):
+        return None
+
+def _calculate_bounding_box_cropdetect(crop_keys):
+    min_x = min_w = min_y = min_h = float('inf')
+    max_x = max_w = max_y = max_h = float('-inf')
+    for key in crop_keys:
+        parsed = _parse_crop_string_cropdetect(key)
+        if not parsed:
+            continue
+        w, h, x, y = parsed['w'], parsed['h'], parsed['x'], parsed['y']
+        min_x = min(min_x, x)
+        min_y = min(min_y, y)
+        max_x = max(max_x, x + w)
+        max_y = max(max_y, y + h)
+        min_w = min(min_w, w)
+        min_h = min(min_h, h)
+        max_w = max(max_w, w)
+        max_h = max(max_h, h)
+    if (max_x - min_x) <= 2 and (max_y - min_y) <= 2:
+        return None
+    bounding_crop = f"crop={max_x - min_x}:{max_y - min_y}:{min_x}:{min_y}"
+    return bounding_crop
+
+def _analyze_video_cropdetect(input_file, duration, width, height, num_workers, significant_crop_threshold, min_crop, debug=False):
+    num_tasks = num_workers * 4
+    segment_duration = max(1, duration // num_tasks)
+    tasks = [(i * segment_duration, input_file, width, height) for i in range(num_tasks)]
+    crop_results = []
+    with _multiprocessing_cropdetect.Pool(processes=num_workers) as pool:
+        results_iterator = pool.imap_unordered(_analyze_segment_cropdetect, tasks)
+        for result in results_iterator:
+            crop_results.append(result)
+    all_crops_with_ts = [crop for sublist in crop_results for crop in sublist]
+    all_crop_strings = [item[0] for item in all_crops_with_ts]
+    if not all_crop_strings:
+        return None
+    crop_counts = _Counter_cropdetect(all_crop_strings)
+    clusters = _cluster_crop_values_cropdetect(crop_counts)
+    total_detections = sum(c['count'] for c in clusters)
+    significant_clusters = []
+    for cluster in clusters:
+        percentage = (cluster['count'] / total_detections) * 100
+        if percentage >= significant_crop_threshold:
+            significant_clusters.append(cluster)
+    for cluster in significant_clusters:
+        parsed_crop = _parse_crop_string_cropdetect(cluster['center'])
+        if parsed_crop:
+            _, ar_label = _snap_to_known_ar_cropdetect(
+                parsed_crop['w'], parsed_crop['h'], parsed_crop['x'], parsed_crop['y'], width, height
+            )
+            cluster['ar_label'] = ar_label
+        else:
+            cluster['ar_label'] = None
+    if not significant_clusters:
+        return None
+    elif len(significant_clusters) == 1:
+        dominant_cluster = significant_clusters[0]
+        parsed_crop = _parse_crop_string_cropdetect(dominant_cluster['center'])
+        snapped_crop, ar_label = _snap_to_known_ar_cropdetect(
+            parsed_crop['w'], parsed_crop['h'], parsed_crop['x'], parsed_crop['y'], width, height
+        )
+        parsed_snapped = _parse_crop_string_cropdetect(snapped_crop)
+        if parsed_snapped and parsed_snapped['w'] == width and parsed_snapped['h'] == height:
+            return None
+        else:
+            return snapped_crop
+    else:
+        crop_keys = [c['center'] for c in significant_clusters]
+        bounding_box_crop = _calculate_bounding_box_cropdetect(crop_keys)
+        if bounding_box_crop:
+            parsed_bb = _parse_crop_string_cropdetect(bounding_box_crop)
+            snapped_crop, ar_label = _snap_to_known_ar_cropdetect(
+                parsed_bb['w'], parsed_bb['h'], parsed_bb['x'], parsed_bb['y'], width, height
+            )
+            parsed_snapped = _parse_crop_string_cropdetect(snapped_crop)
+            if parsed_snapped and parsed_snapped['w'] == width and parsed_snapped['h'] == height:
+                return None
+            else:
+                return snapped_crop
+        else:
+            return None
+
+def detect_autocrop_filter(input_file, significant_crop_threshold=5.0, min_crop=10, debug=False):
+    if not _check_prerequisites_cropdetect():
+        return None
+    try:
+        probe_duration_args = [
+            'ffprobe', '-v', 'error', '-show_entries', 'format=duration', '-of', 'default=noprint_wrappers=1:nokey=1',
+            input_file
+        ]
+        duration_str = subprocess.check_output(probe_duration_args, stderr=subprocess.STDOUT, text=True)
+        duration = int(float(duration_str))
+        probe_res_args = [
+            'ffprobe', '-v', 'error',
+            '-select_streams', 'v',
+            '-show_entries', 'stream=width,height,disposition',
+            '-of', 'json',
+            input_file
+        ]
+        probe_output = subprocess.check_output(probe_res_args, stderr=subprocess.STDOUT, text=True)
+        streams_data = json.loads(probe_output)
+        video_stream = None
+        for stream in streams_data.get('streams', []):
+            if stream.get('disposition', {}).get('attached_pic', 0) == 0:
+                video_stream = stream
+                break
+        if not video_stream or 'width' not in video_stream or 'height' not in video_stream:
+            return None
+        width = int(video_stream['width'])
+        height = int(video_stream['height'])
+    except Exception:
+        return None
+    return _analyze_video_cropdetect(input_file, duration, width, height, max(1, os.cpu_count() // 2), significant_crop_threshold, min_crop, debug)
+
+def main(no_downmix=False, autocrop=False):
     check_tools()
-    
     current_dir = Path(".")
-    
-    # Check if there are any MKV files to process before creating directories
     files_to_process = sorted(
         f for f in current_dir.glob("*.mkv")
         if not (f.name.endswith(".ut.mkv") or f.name.startswith("temp-") or f.name.startswith("output-"))
     )
-
     if not files_to_process:
         print("No MKV files found to process. Exiting.")
-        return  # Exit without creating directories
-        
-    # Only create directories when we actually have files to process
+        return
     DIR_COMPLETED.mkdir(exist_ok=True, parents=True)
     DIR_ORIGINAL.mkdir(exist_ok=True, parents=True)
     DIR_LOGS.mkdir(exist_ok=True, parents=True)
-
     while True:
         files_to_process = sorted(
             f for f in current_dir.glob("*.mkv")
             if not (f.name.endswith(".ut.mkv") or f.name.startswith("temp-") or f.name.startswith("output-"))
         )
-
         if not files_to_process:
             print("No more .mkv files found to process in the current directory. The script will now exit.")
             break
-
         file_path = files_to_process[0]
-
-        # Setup logging
         log_file_path = DIR_LOGS / f"{file_path.name}.log"
         log_file = open(log_file_path, 'w', encoding='utf-8')
         original_stdout = sys.stdout
         original_stderr = sys.stderr
         sys.stdout = Tee(original_stdout, log_file)
         sys.stderr = Tee(original_stderr, log_file)
-
         try:
             print("-" * shutil.get_terminal_size(fallback=(80, 24)).columns)
             print(f"Starting full processing for: {file_path.name}")
             date = datetime.now()
             input_file_abs = file_path.resolve()
             intermediate_output_file = current_dir / f"output-{file_path.name}"
-            audio_temp_dir = None # Initialize to None
+            audio_temp_dir = None
             created_ut_video_path = None
             created_encoded_video_path = None
-
             try:
-                audio_temp_dir = tempfile.mkdtemp(prefix="tv_audio_") # UUID is not strictly needed for uniqueness
+                audio_temp_dir = tempfile.mkdtemp(prefix="tv_audio_")
                 print(f"Audio temporary directory created at: {audio_temp_dir}")
                 print(f"Analyzing file: {input_file_abs}")
 
@@ -229,7 +437,16 @@ def main(no_downmix=False):
                 ], capture_output=True)
                 media_info = json.loads(mediainfo_json)
 
-                created_ut_video_path, created_encoded_video_path = convert_video(file_path.stem, str(input_file_abs))
+                autocrop_filter = None
+                if autocrop:
+                    print("--- Running autocrop detection ---")
+                    autocrop_filter = detect_autocrop_filter(str(input_file_abs))
+                    if autocrop_filter:
+                        print(f"    - Autocrop filter detected: {autocrop_filter}")
+                    else:
+                        print("    - No crop needed or detected.")
+
+                created_ut_video_path, created_encoded_video_path = convert_video(file_path.stem, str(input_file_abs), autocrop_filter=autocrop_filter)
 
                 print("--- Starting Audio Processing ---")
                 processed_audio_files = []
@@ -352,7 +569,8 @@ def main(no_downmix=False):
 
 if __name__ == "__main__":
     import argparse
-    parser = argparse.ArgumentParser(description="Batch-process MKV files with resumable video encoding and audio downmixing.")
+    parser = argparse.ArgumentParser(description="Batch-process MKV files with resumable video encoding and audio downmixing, with optional autocrop.")
     parser.add_argument("--no-downmix", action="store_true", help="Preserve original audio channel layout.")
+    parser.add_argument("--autocrop", action="store_true", help="Automatically detect and crop black bars from video using cropdetect.")
     args = parser.parse_args()
-    main(no_downmix=args.no_downmix)
+    main(no_downmix=args.no_downmix, autocrop=args.autocrop)