feat: add volume normalization using FFmpeg loudnorm filter

Co-authored-by: aider (gemini/gemini-3.1-pro-preview) <aider@aider.chat>
This commit is contained in:
Tanner
2026-07-28 17:08:17 -06:00
parent 43c59d2c28
commit 4b0774b3fa
+21 -1
View File
@@ -26,6 +26,18 @@ def playback_thread(file_path, state):
stream = container.streams.audio[0]
resampler = av.AudioResampler(format='s16', layout='stereo', rate=48000)
def build_graph():
graph = av.filter.Graph()
src = graph.add_abuffer(template=stream)
loudnorm = graph.add("loudnorm", "I=-16:TP=-1.5:LRA=11")
sink = graph.add("abuffersink")
src.link_to(loudnorm)
loudnorm.link_to(sink)
graph.configure()
return graph
graph = build_graph()
with open(fifo_path, 'wb') as fifo:
iterator = container.decode(stream)
while state.running:
@@ -41,6 +53,7 @@ def playback_thread(file_path, state):
with state.lock:
state.seek_request = None
iterator = container.decode(stream)
graph = build_graph() # Reset filter graph on seek
continue
if paused:
@@ -57,7 +70,14 @@ def playback_thread(file_path, state):
with state.lock:
state.time_pos = float(frame.pts * stream.time_base)
resampled_frames = resampler.resample(frame)
graph.push(frame)
while True:
try:
filtered_frame = graph.pull()
except (av.AVError, BlockingIOError):
break
resampled_frames = resampler.resample(filtered_frame)
for r_frame in resampled_frames:
arr = r_frame.to_ndarray()