tune audio

render.py

@@ -21,6 +21,11 @@ HIDDEN_TAGS = [
     "women with white males"
 ]
 
+# items to not use at all
+BLACKLIST = [
+    'XN'
+]
+
 api = None
 
 def get_api():
@@ -130,7 +135,9 @@ def get_clips(tag):
                 'operator': '&'
             },
             'keys': ['id', 'in', 'out'],
-            'range': [0, 10000]})['data']['items']
+            'range': [0, 90000]})['data']['items']
+        clips = [clip for clip in clips if clip['id'].split('/')[0] not in BLACKLIST]
+
         for clip in clips:
             clip['path'] = get_path(clip['id'].split('/')[0])
             # or use round?
@@ -138,6 +145,7 @@ def get_clips(tag):
             clip['out'] = int(clip['out'] / FRAME_DURATION) * FRAME_DURATION
             clip['duration'] = clip['out'] - clip['in']
             clip['tag'] = tag
+        clips = [clip for clip in clips if clip['duration']]
         CLIPS[tag] = list(sorted(clips, key=lambda c: c['id']))
         with open('CLIPS.json', 'w') as fd:
             json.dump(CLIPS, fd, indent=4, sort_keys=True)
@@ -250,7 +258,8 @@ def sequence(seq, letter):
         n = seq()
         if n == 0:
             blank = {'blank': True, 'duration': position - last_text}
-            result['text'].append(blank)
+            if blank['duration']:
+                result['text'].append(blank)
             n = seq()
             if n == 0:
                 n = 10
@@ -301,6 +310,8 @@ def sequence(seq, letter):
                         and result[track][-1]['duration'] > clip['duration']:
                     result[track][-1]['duration'] -= (position-duration)
                     position = duration
+                    if not result[track][-1]['duration']:
+                        result[track].pop(-1)
                 if position <= duration:
                     result[track].append(clip)
                 else:
@@ -312,10 +323,15 @@ def sequence(seq, letter):
     track = 'vocals'
     if letter in VOCALS:
         position = 0
+        loop = 0
         while position < duration:
             n = seq()
+            # vocals should start after one blank
+            if len(result[track])  and result[track][-1].get('blank'):
+                n = 10
+            # 50 % chance of a silence of up to 5 seconds
             if n < 5:
-                n = seq()
+                n = seq() / 2  # (0 - 5 seconds)
                 position += add_blank(result[track], min(n, duration-position))
             else:
                 n = seq()
@@ -326,11 +342,15 @@ def sequence(seq, letter):
                         and result[track][-1]['duration'] > clip['duration']:
                     result[track][-1]['duration'] -= (position-duration)
                     position = duration
+                    if not result[track][-1]['duration']:
+                        result[track].pop(-1)
                 if position <= duration:
                     result[track].append(clip)
                 else:
                     position -= clip['duration']
-                    break
+                    if duration - position < 10 or loop > 10:
+                        break
+                    loop += 1
         if position < duration:
             position += add_blank(result[track], duration - position)
     '''
@@ -380,6 +400,7 @@ def sequence(seq, letter):
 
 
 if __name__ == '__main__':
+    encode = len(sys.argv) < 2 or sys.argv[1] != 'json'
     for n in range(10):
         seq = random(n * 1000)
         #for letter in ('T', 'W'):
@@ -400,5 +421,7 @@ if __name__ == '__main__':
             if current != old:
                 with open(tjson, 'w') as fd:
                     fd.write(current)
-                subprocess.call(['./render_mlt.py', tjson])
+            if encode:
-                #subprocess.call(['./render_audio.py', tjson])
+                if current != old or os.path.getmtime(tjson) < os.path.getmtime('render_mlt.py'):
+                    subprocess.call(['./render_mlt.py', tjson, 'encode'])
+                    #subprocess.call(['./render_audio.py', tjson])

render_mlt.py

@@ -5,6 +5,7 @@ import sys
 import json
 import subprocess
 
+import ox
 import mlt
 from PyQt5 import QtWidgets
 
@@ -14,7 +15,6 @@ app = QtWidgets.QApplication(sys.argv)
 #mlt.mlt_log_set_level(40) # verbose
 mlt.Factory.init()
 
-
 tractor = mlt.Tractor()
 tractor.mark_in = -1
 tractor.mark_out = -1
@@ -22,6 +22,11 @@ tractor.mark_out = -1
 multitrack = tractor.multitrack()
 
 source = sys.argv[1]
+if len(sys.argv) > 2:
+    encode = sys.argv[2] == 'encode'
+else:
+    encode = False
+
 target = source.replace('.json', '.xml')
 target_audio = source.replace('.json', '.audio.xml')
 target_vocals = source.replace('.json', '.vocals.xml')
@@ -52,9 +57,25 @@ def add_color(playlist, color, duration):
 def add_clip(playlist, file_, in_, duration):
     if not isinstance(file_, str):
         file_ = file_.encode('utf-8')
+    '''
+    info = ox.avinfo(file_)
+    tractor = mlt.Tractor()
+    tracks = tractor.multitrack()
+    video = mlt.Playlist()
+    '''
     clip = mlt.Producer(profile, file_)
     clip.set_in_and_out(in_, in_+duration-1)
     playlist.append(clip)
+    '''
+    video.append(clip)
+    tracks.connect(video, 0)
+    if not not info.get('audio'):
+        audio = mlt.Playlist()
+        add_silence(audio, duration)
+        tracks.connect(audio, 1)
+    #tracks.set_in_and_out(in_, in_+duration-1)
+    playlist.append(tractor)
+    '''
 
 def add_audio_clip(playlist, file_, duration):
     in_ = 0
@@ -82,9 +103,11 @@ def add_text(playlist, value, length):
 
 
 for clip in data['clips']:
+    frames = int(clip['duration'] * fps)
+    if not frames:
+        continue
     if clip.get('black'):
         # fixme seconds to fps! duration fame etc!!
-        frames = int(clip['duration'] * fps)
         add_color(video, 'black', frames)
     else:
         #print(clip['duration'], clip['path'])
@@ -93,20 +116,22 @@ for clip in data['clips']:
             sys.exit(1)
         # fixme seconds to fps!
         in_ = int(clip['in'] * fps)
-        frames = int(clip['duration'] * fps)
         add_clip(video, clip['path'], in_, frames)
 add_color(video, 'black', 60)
 
 for clip in data['text']:
+    frames = int(clip['duration'] * fps)
+    if not frames:
+        continue
     if clip.get('blank'):
-        frames = int(clip['duration'] * fps)
         add_blank(overlay, frames)
     else:
-        frames = int(clip['duration'] * fps)
         add_text(overlay, clip['text'], frames)
 
 for clip in data['music']:
     frames = int(clip['duration'] * fps)
+    if not frames:
+        continue
     if clip.get('blank'):
         add_silence(music, frames)
     else:
@@ -114,6 +139,8 @@ for clip in data['music']:
 
 for clip in data['vocals']:
     frames = int(clip['duration'] * fps)
+    if not frames:
+        continue
     if clip.get('blank'):
         add_silence(vocals, frames)
     else:
@@ -138,7 +165,7 @@ composite = mlt.Transition(profile, "composite")
 tractor.plant_transition(composite)
 
 volume = mlt.Filter(profile, "volume")
-volume.set("gain", '0.1')
+volume.set("gain", '0.12')
 tractor.plant_filter(volume)
 
 def mix_audio_tracks(a, b, ratio):
@@ -152,6 +179,8 @@ def mix_audio_tracks(a, b, ratio):
     mix = mlt.Transition(profile, "mix")
     mix.set("start", ratio)
     mix.set("end", ratio)
+    #mix.set("always_active", 1)
+    #mix.set("combine", 1)
     tractor.plant_transition(mix)
     return tractor
 
@@ -164,9 +193,15 @@ consumer.start()
 # mix drones
 drones = mix_audio_tracks(drones0, drones1, 0.5)
 
+
 # mix drones + music
-mtractor = mix_audio_tracks(drones, music, 0.20)
+mtractor = mix_audio_tracks(drones, music, 0.3)
-atractor = mix_audio_tracks(vocals, mtractor, 0.20)
+norm = mlt.Filter(profile, "volume")
+norm.set("gain", "-12dB")
+mtractor.plant_filter(norm)
+
+# background and vocals
+atractor = mix_audio_tracks(vocals, mtractor, 0.4)
 
 consumer = mlt.Consumer(profile, 'xml', target_audio)
 consumer.connect(atractor)
@@ -176,14 +211,17 @@ target_audio_wav = target_audio + '.wav'
 subprocess.call([
     'qmelt', target_audio, '-consumer', 'avformat:' + target_audio_wav,
 ])
-audiomix = mlt.Producer(profile, target_audio_wav)
+
+audiomix = mlt.Playlist()
+duration = sum(clip['duration'] for clip in data['clips'])
+add_audio_clip(audiomix, target_audio_wav, int(duration * fps))
 
 # mix vocals and music
 #atractor = mix_audio_tracks(vocals, mtractor, 0.20)
 
 # mix video + audio
 #dtractor = mix_audio_tracks(atractor, tractor, 0.5)
-dtractor = mix_audio_tracks(audiomix, tractor, 0.5)
+dtractor = mix_audio_tracks(audiomix, tractor, 0.29)
 
 
 output = mlt.Tractor()
@@ -193,7 +231,8 @@ output_tracks = output.multitrack()
 output_tracks.connect(dtractor, 0)
 
 norm = mlt.Filter(profile, "volume")
-norm.set("gain", "-6dB")
+#norm.set("gain", "-6dB")
+norm.set("gain", "3dB")
 output.plant_filter(norm)
 
 consumer = 'xml'
@@ -201,3 +240,6 @@ consumer = mlt.Consumer(profile, consumer, target)
 consumer.connect(output)
 #consumer.set("real_time", -2)
 consumer.start()
+
+if encode:
+    subprocess.call(['./encode.py', target])