more cleanup

src/editors/sentiment/editor.py

@@ -1,4 +1,7 @@
 import whisper
+import json
+from pathlib import Path
+import tempfile
 import numpy as np
 import structlog
 
@@ -19,10 +22,22 @@ class TextGlob:
 class SentimentEditor:
     def __init__(self, video_path, audio_path, params):
         self.logger = structlog.get_logger("sentiment")
-        self.logger.info("loading whisper model", size=params["model_size"])
-        self.model = whisper.load_model(params["model_size"])
-        self.logger.info("transcribing audio", path=audio_path)
-        self.result = self.model.transcribe(audio_path)
+        tempdir = tempfile.gettempdir()
+        dest_location = f"{tempdir}/{params['temp_file_name']}-{params['model_size']}-sentiment.json"
+        if not Path(dest_location).is_file():
+            self.logger.info("loading whisper model", size=params["model_size"])
+            self.model = whisper.load_model(params["model_size"])
+            self.logger.info("transcribing audio", path=audio_path)
+            self.result = self.model.transcribe(audio_path)
+
+            with open(dest_location, 'w') as fp:
+                json.dump(self.result, fp)
+        else:
+            self.logger.info("cached transcription found", path=dest_location)
+
+        with open(dest_location, 'r') as f:
+          self.result = json.load(f)
+
         self.segments = []
         for segment in self.result['segments']:
             self.segments.append(TextGlob(segment['start'], segment['end'], segment['text'], 0))
@@ -42,7 +57,7 @@ class SentimentEditor:
     def edit(self, large_window, small_window, params):
         end_time = self.segments[-1].stop
         window_factor = len(self.sentiments) / end_time
-        long_ma = np_moving_average(self.squared_subsample, large_window * window_factor)
-        short_ma = np_moving_average(self.squared_subsample, small_window * window_factor)
+        long_ma = np_moving_average(self.sentiments, large_window)
+        short_ma = np_moving_average(self.sentiments, small_window)
         highlights = find_moving_average_highlights(short_ma, long_ma, 1.0 / window_factor)
         return highlights, large_window, small_window

src/mediautils/video.py

@@ -9,5 +9,8 @@ def get_subclips(source_video_path, moments):
 
 def render_moments(moments, input_video_path, output_path):
         clips, vid = get_subclips(input_video_path, moments)
-        to_render = mp.concatenate_videoclips(clips, logger=None)
+        to_render = mp.concatenate_videoclips(clips)
         to_render.write_videofile(output_path, logger=None)
+
+def filter_moments(moments, min_length, max_length):
+    return [m for m in moments if m.get_duration() > min_length and m.get_duration() < max_length]