GameSentenceMiner 2.7.14__py3-none-any.whl → 2.7.16__py3-none-any.whl

GameSentenceMiner/ocr/owocr_area_selector.py

@@ -283,10 +283,10 @@ class ScreenSelector:
                 win_w = window_geom_to_save['width']
                 win_h = window_geom_to_save['height']
                 # Basic check for valid dimensions needed for percentage calculation
-                if win_w > 0 and win_h > 0 and win_h > 0 and win_t > 0:
+                if win_w > 0 and win_h > 0:
                     save_coord_system = COORD_SYSTEM_PERCENTAGE
-                    win_l = max(0, window_geom_to_save['left'])
-                    win_t = max(0, window_geom_to_save['top'])
+                    win_l = window_geom_to_save['left']
+                    win_t = window_geom_to_save['top']
                     print(f"Saving using coordinate system: {save_coord_system} relative to {window_geom_to_save}")
                 else:
                     print(
@@ -306,14 +306,14 @@ class ScreenSelector:
                 coords_to_save = []
 
                 # --- Convert absolute pixels to the chosen system ---
-                if save_coord_system == COORD_SYSTEM_PERCENTAGE and window_geom_to_save and 0 <= win_l < monitor_dict['left'] and 0 <= win_t < monitor_dict['top']:
+                if save_coord_system == COORD_SYSTEM_PERCENTAGE and window_geom_to_save:
                     # Calculate percentages (handle potential float precision issues if necessary)
                     x_pct = (x_abs - win_l) / win_w
                     y_pct = (y_abs - win_t) / win_h
                     w_pct = w_abs / win_w
                     h_pct = h_abs / win_h
                     # Round percentages slightly to avoid overly long floats? Optional.
-                    # precision = 6
+                    # precision = 6+
                     # coords_to_save = [round(x_pct, precision), round(y_pct, precision), round(w_pct, precision), round(h_pct, precision)]
                     coords_to_save = [x_pct, y_pct, w_pct, h_pct]
                 else:
@@ -861,8 +861,8 @@ if __name__ == "__main__":
         # Example: uncomment below to target Calculator on Windows by default if no arg given
         # if sys.platform == "win32": target_window_title = "Calculator"
 
-    if not target_window_title:
-        target_window_title = get_ocr_config().window
+    # if not target_window_title:
+    #     target_window_title = get_ocr_config().window
 
     # Get the selection result
     selection_result = get_screen_selection(target_window_title)

GameSentenceMiner/ocr/owocr_helper.py

@@ -84,7 +84,8 @@ def get_ocr_config() -> OCRConfig:
     scene = util.sanitize_filename(obs.get_current_scene())
     config_path = ocr_config_dir / f"{scene}.json"
     if not config_path.exists():
-        raise Exception(f"No config file found at {config_path}.")
+        config_path.touch()
+        return
     try:
         with open(config_path, 'r', encoding="utf-8") as f:
             config_data = json.load(f)
@@ -107,20 +108,21 @@ def get_ocr_config() -> OCRConfig:
                         "coordinates": rect,
                         "is_excluded": False
                     })
-                for rect in config_data['excluded_rectangles']:
-                    new_rectangles.append({
-                        "monitor": {
-                            "left": default_monitor["left"],
-                            "top": default_monitor["top"],
-                            "width": default_monitor["width"],
-                            "height": default_monitor["height"],
-                            "index": 0  # Assuming single monitor for old config
-                        },
-                        "coordinates": rect,
-                        "is_excluded": True
-                    })
+                if 'excluded_rectangles' in config_data:
+                    for rect in config_data['excluded_rectangles']:
+                        new_rectangles.append({
+                            "monitor": {
+                                "left": default_monitor["left"],
+                                "top": default_monitor["top"],
+                                "width": default_monitor["width"],
+                                "height": default_monitor["height"],
+                                "index": 0  # Assuming single monitor for old config
+                            },
+                            "coordinates": rect,
+                            "is_excluded": True
+                        })
             new_config_data = {"scene": config_data.get("scene", scene), "window": config_data.get("window", None),
-                               "rectangles": new_rectangles}
+                               "rectangles": new_rectangles, "coordinate_system": "absolute"}
             with open(config_path, 'w', encoding="utf-8") as f:
                 json.dump(new_config_data, f, indent=4)
             return OCRConfig.from_dict(new_config_data)
@@ -231,32 +233,45 @@ def do_second_ocr(ocr1_text, rectangle_index, time, img):
 last_oneocr_results_to_check = {}  # Store last OCR result for each rectangle
 last_oneocr_times = {}    # Store last OCR time for each rectangle
 text_stable_start_times = {} # Store the start time when text becomes stable for each rectangle
+orig_text_results = {} # Store original text results for each rectangle
 TEXT_APPEARENCE_DELAY = get_ocr_scan_rate() * 1000 + 500  # Adjust as needed
 
-def text_callback(text, rectangle_index, time, img=None):
-    global twopassocr, ocr2, last_oneocr_results_to_check, last_oneocr_times, text_stable_start_times
+def text_callback(text, orig_text, rectangle_index, time, img=None):
+    global twopassocr, ocr2, last_oneocr_results_to_check, last_oneocr_times, text_stable_start_times, orig_text_results
+    orig_text_string = ''.join([item for item in orig_text if item is not None]) if orig_text else ""
 
     current_time = time if time else datetime.now()
 
     previous_text = last_oneocr_results_to_check.get(rectangle_index, "").strip()
+    previous_orig_text = orig_text_results.get(rectangle_index, "").strip()
+
+    # print(previous_orig_text)
+    # if orig_text:
+    #     print(orig_text_string)
 
     if not text:
         if previous_text:
             if rectangle_index in text_stable_start_times:
                 stable_time = text_stable_start_times[rectangle_index]
+                previous_result = last_ocr1_results[rectangle_index]
+                if previous_result and fuzz.ratio(previous_result, previous_text) >= 80:
+                    logger.info("Seems like the same text, not " + "doing second OCR" if twopassocr else "sending")
+                    del last_oneocr_results_to_check[rectangle_index]
+                    return
+                if previous_orig_text and fuzz.ratio(orig_text_string, previous_orig_text) >= 80:
+                    logger.info("Seems like Text we already sent, not doing anything.")
+                    del last_oneocr_results_to_check[rectangle_index]
+                    return
+                orig_text_results[rectangle_index] = orig_text_string
                 if twopassocr:
                     do_second_ocr(previous_text, rectangle_index, time, img)
                 else:
-                    previous_result = last_ocr1_results[rectangle_index]
-                    if previous_result and fuzz.ratio(previous_result, previous_text) >= 80:
-                        logger.info("Seems like the same text, not sending")
-                        return
                     if get_config().advanced.ocr_sends_to_clipboard:
                         import pyperclip
                         pyperclip.copy(text)
                     websocket_server_thread.send_text(previous_text, stable_time)
                     img.save(os.path.join(get_app_directory(), "temp", "last_successful_ocr.png"))
-                    last_ocr1_results[rectangle_index] = previous_text
+                last_ocr1_results[rectangle_index] = previous_text
                 del text_stable_start_times[rectangle_index]
             del last_oneocr_results_to_check[rectangle_index]
             return
@@ -285,11 +300,13 @@ done = False
 
 def run_oneocr(ocr_config: OCRConfig, i, area=False):
     global done
-    rect_config = ocr_config.rectangles[i]
-    coords = rect_config.coordinates
-    monitor_config = rect_config.monitor
+    screen_area = None
+    if ocr_config.rectangles:
+        rect_config = ocr_config.rectangles[i]
+        coords = rect_config.coordinates
+        monitor_config = rect_config.monitor
+        screen_area = ",".join(str(c) for c in coords) if area else None
     exclusions = list(rect.coordinates for rect in list(filter(lambda x: x.is_excluded, ocr_config.rectangles)))
-    screen_area = ",".join(str(c) for c in coords) if area else None
     run.run(read_from="screencapture", write_to="callback",
             screen_capture_area=screen_area,
             # screen_capture_monitor=monitor_config['index'],
@@ -326,9 +343,9 @@ if __name__ == "__main__":
     logger.info(f"Received arguments: ocr1={ocr1}, ocr2={ocr2}, twopassocr={twopassocr}")
     global ocr_config
     ocr_config: OCRConfig = get_ocr_config()
-
+    print(ocr_config)
     logger.info(f"Starting OCR with configuration: Window: {ocr_config.window}, Rectangles: {len(ocr_config.rectangles)}, Engine 1: {ocr1}, Engine 2: {ocr2}, Two-pass OCR: {twopassocr}")
-    if ocr_config and ocr_config.rectangles:
+    if ocr_config:
         rectangles = list(filter(lambda rect: not rect.is_excluded, ocr_config.rectangles))
         last_ocr1_results = [""] * len(rectangles) if rectangles else [""]
         last_ocr2_results = [""] * len(rectangles) if rectangles else [""]

GameSentenceMiner/owocr/owocr/ocr.py

@@ -759,7 +759,9 @@ class OneOCR:
 
         if sys.platform == 'win32':
             try:
-                res = self.model.recognize_pil(img)['text']
+                ocr_resp = self.model.recognize_pil(img)
+                # print(json.dumps(ocr_resp))
+                res = ocr_resp['text']
             except RuntimeError as e:
                 return (False, e)
         else:
@@ -773,6 +775,7 @@ class OneOCR:
             if res.status_code != 200:
                 return (False, 'Unknown error!')
 
+
             res = res.json()['text']
 
         x = (True, res)
@@ -1007,9 +1010,9 @@ class GeminiOCR:
         try:
             import google.generativeai as genai
             if isinstance(img_or_path, str) or isinstance(img_or_path, Path):
-                img = Image.open(img_or_path).convert("RGB")
+                img = Image.open(img_or_path)
             elif isinstance(img_or_path, Image.Image):
-                img = img_or_path.convert("RGB")
+                img = img_or_path
             else:
                 raise ValueError(f'img_or_path must be a path or PIL.Image, instead got: {img_or_path}')
 
@@ -1023,11 +1026,11 @@ class GeminiOCR:
                         {
                             'inline_data': {
                                 'mime_type': 'image/png',
-                                'data': base64.b64encode(img_bytes).decode('utf-8')
+                                'data': img_bytes
                             }
                         },
                         {
-                            'text': 'As Quick as Possible, Give me the text from this image, no other output. If there is no text, return nothing.'
+                            'text': 'Analyze the image. Extract text *only* from within dialogue boxes (speech bubbles or panels containing character dialogue). From the extracted dialogue text, filter out any furigana. Ignore and do not include any text found outside of dialogue boxes, including character names, speaker labels, or sound effects. Return *only* the filtered dialogue text. If no text is found within dialogue boxes after applying filters, return nothing. Do not include any other output, formatting markers, or commentary.'
                         }
                     ]
                 }
@@ -1044,11 +1047,4 @@ class GeminiOCR:
             return (False, f'Gemini API request failed: {e}')
 
     def _preprocess(self, img):
-        try:
-            from io import BytesIO
-            img_io = BytesIO()
-            img.save(img_io, 'PNG')  # Save as PNG
-            return img_io.getvalue()
-        except Exception as e:
-            logger.error(f'Error preprocessing image for Gemini: {e}')
-            return None
+        return pil_image_to_bytes(img, png_compression=1)

GameSentenceMiner/owocr/owocr/run.py

@@ -25,6 +25,10 @@ import psutil
 
 import inspect
 from .ocr import *
+try:
+    from .secret import *
+except ImportError:
+    pass
 from .config import Config
 from .screen_coordinate_picker import get_screen_selection
 from ...configuration import get_temporary_directory
@@ -571,7 +575,7 @@ def process_and_write_results(img_or_path, write_to, notifications, last_result,
         elif write_to == 'clipboard':
             pyperclipfix.copy(text)
         elif write_to == "callback":
-            txt_callback(text, rectangle, start_time, img_or_path)
+            txt_callback(text, orig_text, rectangle, start_time, img_or_path)
         elif write_to:
             with Path(write_to).open('a', encoding='utf-8') as f:
                 f.write(text + '\n')
@@ -670,7 +674,10 @@ def run(read_from=None,
         for config_engine in config.get_general('engines').split(','):
             config_engines.append(config_engine.strip().lower())
 
-    for _,engine_class in sorted(inspect.getmembers(sys.modules[__name__], lambda x: hasattr(x, '__module__') and x.__module__ and __package__ + '.ocr' in x.__module__ and inspect.isclass(x))):
+    for _, engine_class in sorted(inspect.getmembers(sys.modules[__name__],
+                                                     lambda x: hasattr(x, '__module__') and x.__module__ and (
+                                                             __package__ + '.ocr' in x.__module__ or __package__ + '.secret' in x.__module__) and inspect.isclass(
+                                                             x))):
         if len(config_engines) == 0 or engine_class.name in config_engines:
             if config.get_engine(engine_class.name) == None:
                 engine_instance = engine_class()
@@ -921,7 +928,7 @@ def run(read_from=None,
     logger.opt(ansi=True).info(f"Reading from {read_from_readable}, writing to {write_to_readable} using <{engine_color}>{engine_instances[engine_index].readable_name}</{engine_color}>{' (paused)' if paused else ''}")
 
     while not terminated and not stop_running_flag:
-        start_time = datetime.datetime.now()
+        start_time = datetime.now()
         if read_from == 'websocket':
             while True:
                 try:

{gamesentenceminer-2.7.14.dist-info → gamesentenceminer-2.7.16.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: GameSentenceMiner
-Version: 2.7.14
+Version: 2.7.16
 Summary: A tool for mining sentences from games. Update: Multi-Line Mining! Fixed!
 Author-email: Beangate <bpwhelan95@gmail.com>
 License: MIT License

{gamesentenceminer-2.7.14.dist-info → gamesentenceminer-2.7.16.dist-info}/RECORD

@@ -23,22 +23,22 @@ GameSentenceMiner/downloader/download_tools.py,sha256=mI1u_FGBmBqDIpCH3jOv8DOoZ3
 GameSentenceMiner/ocr/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 GameSentenceMiner/ocr/gsm_ocr_config.py,sha256=zagsB4UD9mmZX_r6dFBCXZqdDa0XGk-RvIqbKoPB9lQ,1932
 GameSentenceMiner/ocr/ocrconfig.py,sha256=hTROOZ3On2HngXKxwQFZvnr5AxlmlMV0mPxv-F3NbMg,6476
-GameSentenceMiner/ocr/owocr_area_selector.py,sha256=bCgusYXe9ibCsf56PlU301aNfDA2PDKasi78ox0IGbk,46856
-GameSentenceMiner/ocr/owocr_helper.py,sha256=Rd1wcVtJy0N6ySn1p7-08hLYA9iPkts12zGJESoEewI,15126
+GameSentenceMiner/ocr/owocr_area_selector.py,sha256=bwlvvM_SwRHzwbZ3GSQfxGHT0ASy3rMxB5DQ7RhVZkQ,46742
+GameSentenceMiner/ocr/owocr_helper.py,sha256=wL6EjjFTU6WJu_1UdY0g1dl0JhLweO54YnAY9fOPjaQ,16117
 GameSentenceMiner/owocr/owocr/__init__.py,sha256=opjBOyGGyEqZCE6YdZPnyt7nVfiwyELHsXA0jAsjm14,25
 GameSentenceMiner/owocr/owocr/__main__.py,sha256=r8MI6RAmbkTWqOJ59uvXoDS7CSw5jX5war9ULGWELrA,128
 GameSentenceMiner/owocr/owocr/config.py,sha256=738QCJHEWpFhMh966plOcXYWwcshSiRsxjjIwldeTtI,7461
 GameSentenceMiner/owocr/owocr/lens_betterproto.py,sha256=oNoISsPilVVRBBPVDtb4-roJtAhp8ZAuFTci3TGXtMc,39141
-GameSentenceMiner/owocr/owocr/ocr.py,sha256=t0kU2GQyW0gf0NGqaYiOO7SjYgX8mQXLaNKJ8Eup6mg,39704
-GameSentenceMiner/owocr/owocr/run.py,sha256=0MGrhO6HoLNF1JpYcl-tS6SnXNM5zu9y0oSZDrAcw5k,47499
+GameSentenceMiner/owocr/owocr/ocr.py,sha256=n24Xg8Z8dbcgLpq1u4d22z3tLV1evmf0dK3-Xocv3vs,39878
+GameSentenceMiner/owocr/owocr/run.py,sha256=pEHxot24yRPvbeE3fmgWMcIILBQfUiSQ3sXjw0LJOF0,47791
 GameSentenceMiner/owocr/owocr/screen_coordinate_picker.py,sha256=fjJ3CSXLti3WboGPpmsa7MWOwIXsfpHC8N4zKahGGY0,3346
 GameSentenceMiner/vad/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 GameSentenceMiner/vad/silero_trim.py,sha256=ULf3zwS-JMsY82cKF7gZxREHw8L6lgpWF2U1YqgE9Oc,1681
 GameSentenceMiner/vad/vosk_helper.py,sha256=125X8C9NxFPlWWpoNsbOnEqKx8RCjXN109zNx_QXhyg,6070
 GameSentenceMiner/vad/whisper_helper.py,sha256=JJ-iltCh813XdjyEw0Wn5DaErf6PDqfH0Efu1Md8cIY,3543
-gamesentenceminer-2.7.14.dist-info/licenses/LICENSE,sha256=OXLcl0T2SZ8Pmy2_dmlvKuetivmyPd5m1q-Gyd-zaYY,35149
-gamesentenceminer-2.7.14.dist-info/METADATA,sha256=KUj0VZC1ZOiywqGwMSjz4No-9EKoJ4g-KgVgoeyDRsU,5892
-gamesentenceminer-2.7.14.dist-info/WHEEL,sha256=pxyMxgL8-pra_rKaQ4drOZAegBVuX-G_4nRHjjgWbmo,91
-gamesentenceminer-2.7.14.dist-info/entry_points.txt,sha256=2APEP25DbfjSxGeHtwBstMH8mulVhLkqF_b9bqzU6vQ,65
-gamesentenceminer-2.7.14.dist-info/top_level.txt,sha256=V1hUY6xVSyUEohb0uDoN4UIE6rUZ_JYx8yMyPGX4PgQ,18
-gamesentenceminer-2.7.14.dist-info/RECORD,,
+gamesentenceminer-2.7.16.dist-info/licenses/LICENSE,sha256=OXLcl0T2SZ8Pmy2_dmlvKuetivmyPd5m1q-Gyd-zaYY,35149
+gamesentenceminer-2.7.16.dist-info/METADATA,sha256=FnLkWirfE9GWkCtHhyegdVXNR9tvpwoIqPoUQBQElAw,5892
+gamesentenceminer-2.7.16.dist-info/WHEEL,sha256=pxyMxgL8-pra_rKaQ4drOZAegBVuX-G_4nRHjjgWbmo,91
+gamesentenceminer-2.7.16.dist-info/entry_points.txt,sha256=2APEP25DbfjSxGeHtwBstMH8mulVhLkqF_b9bqzU6vQ,65
+gamesentenceminer-2.7.16.dist-info/top_level.txt,sha256=V1hUY6xVSyUEohb0uDoN4UIE6rUZ_JYx8yMyPGX4PgQ,18
+gamesentenceminer-2.7.16.dist-info/RECORD,,