def speaker_wav(wav_fn, alignment_path, speaker):
basename = ".".join(os.path.basename(wav_fn).split('.')[:-1])
alignment = json.load(open(alignment_path, 'r'))["words"]
c = C.Composition(channels=1)
t = C.Track(wav_fn, speaker)
c.add_track(t)
score_loc = 0.0
start = None
end = None
for word in alignment:
if "speaker" in word:
if word["speaker"] == speaker:
if start is None:
start = word["start"]
end = word["end"]
else:
if start is not None:
seg = C.Segment(t, start, start, end - start)
c.add_segment(seg)
start = None
end = None
if start is not None:
seg = C.Segment(t, start, start, word["end"] - start)
c.add_segment(seg)
out_fn = "static/speechtracks/%s-%s" % (basename, speaker)
c.export(
min_length=t.duration,
filename=out_fn,
channels=1,
filetype='wav',
samplerate=t.samplerate,
separate_tracks=False)
return out_fn + ".wav"
def find_loop(self, start_time, end_time):
graph = self.graph
# create a subgraph with just the nodes in the time frame
span_nodes = []
for n in graph.nodes(data=True):
if float(n[0]) > start_time and float(n[0]) < end_time:
span_nodes.append(n[0])
subgraph = graph.subgraph(span_nodes)
cycles = nx.simple_cycles(subgraph)
if len(cycles) == 0:
print "There are no loops in this time span. Alas!"
return
longest_cycle = max(cycles, key=lambda x: len(x))
print "Longest cycle:", longest_cycle
# compose the audio
wav_fn = ".".join(self.filename.split('.')[:-1]) + '.wav'
try:
with open(wav_fn) as f:
pass
except IOError as e:
# convert to wav
subprocess.call('lame --decode "%s"' % self.filename, shell=True)
track = C.Track(wav_fn, "song to remix")
c = C.Composition(tracks=[track])
score_loc = 0.0
for i, node in enumerate(longest_cycle[:-1]):
for edge in subgraph.edges_iter(nbunch=[node], data=True):
if edge[1] == longest_cycle[i + 1]:
segment = C.Segment(track, score_loc, node,
edge[2]["duration"])
c.add_score_segment(segment)
score_loc += edge[2]["duration"]
break
c.output_score(adjust_dynamics=False,
filename="boom",
channels=2,
filetype='wav',
separate_tracks=False)
c = C.Composition(channels=1)
t = C.Track(wav_fn, speaker)
c.add_track(t)
score_loc = 0.0
start = None
end = None
for word in transcript:
if "speaker" in word:
if word["speaker"] == speaker:
if start is None:
start = word["start"]
end = word["end"]
else:
if start is not None:
seg = C.Segment(t, start, start, end - start)
c.add_score_segment(seg)
start = None
end = None
if start is not None:
seg = C.Segment(t, start, start, word["end"] - start)
c.add_score_segment(seg)
c.output_score(adjust_dynamics=False,
min_length=t.duration,
filename="static/speechtracks/%s-%s" % (basename, speaker),
channels=1,
filetype='wav',
samplerate=t.samplerate,
separate_tracks=False)