Синхронизация аудио/видео в декодировании pyav ⇐ Python

[Anonymous]

I am making a simple video player with pyav, and using the container.decode(video=0, audio=0) method to decode both audio and video frames, then syncing them using PIL and pyaudio.

Here is the summary code/method I used:

import av import pyaudio from PIL import ImageTk import tkinter as tk ... with av.open(path, "r") as self._container: audio_stream = self._container.streams.audio[0] p = pyaudio.PyAudio() audio_stream = p.open(format=pyaudio.paFloat32, channels=audio_stream.channels, rate=audio_stream.rate, output=True) while True: try: frame = next(self._container.decode(video=0, audio=0)) # decode both audio and video simultaneously if 'Video' in repr(frame): # show the video frame self._current_img = frame.to_image() # converts the video frame to pil image self._current_imgtk = ImageTk.PhotoImage(self._current_img) self.label.config(image=self._current_imgtk) # display the image in a tk label else: # play the audio frame audio_data = frame.to_ndarray().astype('float32') interleaved_data = audio_data.T.flatten().tobytes() audio_stream.write(interleaved_data) except (StopIteration, av.error.EOFError, tk.TclError): break self._container = None audio_stream.stop_stream() audio_stream.close() p.terminate() ... The video/audio is properly synced when decoding is done like this (according to pts):

video_frame audio_frame audio_frame video_frame audio_frame audio_frame ... But not all videos are decoded like this, instead a continues series is decoded and not synced properly:

video_frame video_frame video_frame video_frame audio_frame audio_frame audio_frame audio_frame audio_frame audio_frame video_frame video_frame video_frame video_frame ... it just stutters between audio and video frames

Is there any method to decode both frames evenly for all videos? Or at least provide any similar project of video player using pyav like this.

I am actually working on this project: https://github.com/Akascape/tkVideoPlayer

Thank you


Источник: https://stackoverflow.com/questions/780 ... v-decoding