Recherche avancée

Sur d’autres sites (4336)

• avformat_close_input memory leak ?

  20 janvier 2021, par Keen Jackdaw

  I developed an app to push live stream with ffmpeg. When I checked the app with leaks --atExit -- <the app="app"></the> (I'm on mac), I found some memory leak with AVFormatContext.

  


  The minimized code are provided below :

  


  #include <iostream>&#xA;&#xA;extern "C" {&#xA;#include <libavcodec></libavcodec>avcodec.h>&#xA;#include <libavformat></libavformat>avformat.h>&#xA;#include <libavdevice></libavdevice>avdevice.h>&#xA;}&#xA;&#xA;void foo() {&#xA;    avdevice_register_all();&#xA;&#xA;    AVFormatContext *avInputFormatContext = avformat_alloc_context();&#xA;    AVInputFormat *avInputFormat = av_find_input_format("avfoundation");&#xA;    std::cout &lt;&lt; "open input" &lt;&lt; std::endl;&#xA;    int ret = avformat_open_input(&amp;avInputFormatContext, "Capture screen 0", avInputFormat, nullptr);&#xA;    if (ret &lt; 0) { std::cout &lt;&lt; "open input failed: " &lt;&lt; ret &lt;&lt; std::endl; return;}&#xA;&#xA;    avformat_close_input(&amp;avInputFormatContext);&#xA;&#xA;}&#xA;&#xA;int main() {&#xA;    foo();&#xA;    return 0;&#xA;}&#xA;&#xA;</iostream>

  


  The output is

  


  Process:         ffmpegtest [87726]
Path:            /Users/USER/*/ffmpegtest
Load Address:    0x10a752000
Identifier:      ffmpegtest
Version:         ???
Code Type:       X86-64
Platform:        macOS
Parent Process:  leaks [87725]

Date/Time:       2021-01-20 15:44:57.533 +0800
Launch Time:     2021-01-20 15:44:55.760 +0800
OS Version:      macOS 11.1 (20C69)
Report Version:  7
Analysis Tool:   /Applications/Xcode.app/Contents/Developer/usr/bin/leaks
Analysis Tool Version:  Xcode 12.3 (12C33)

Physical footprint:         9.9M
Physical footprint (peak):  10.6M
----

leaks Report Version: 4.0
Process 87726: 14143 nodes malloced for 2638 KB
Process 87726: 1 leak for 32 total leaked bytes.

    1 (32 bytes) ROOT LEAK: 0x7f8c61e1b040 [32]  length: 16  "Capture screen 0"


  


  Did I miss something ?

  


• FFmpeg from Python has Corrupted Output

  14 juillet 2019, par Mazyod

  In a python script, I have :

  os.system(ff_cmd)
  
  # also tried Popen(ff_cmd, shell=True).wait()

  Then, I ran that same ff_cmd from terminal directly. Here are the results :

  # ff_cmd
  
  ffmpeg -i "114006.mp3" -acodec pcm_s16le -ar 16000 -ac 1 "114-006-a4dec52a.wav"
  
  
  
  # after running from python
  
  % file 114-006-a4dec52a.wav
  
  114-006-a4dec52a.wav: Audio file with ID3 version 2.4.0, contains:MPEG ADTS, layer III, v2,  40 kbps, 16 kHz, Monaural
  
  
  
  # after running from terminal
  
  % file 114-006-a4dec52a.wav
  
  114-006-a4dec52a.wav: RIFF (little-endian) data, WAVE audio, Microsoft PCM, 16 bit, mono 16000 Hz

  Can anyone please explain why in the world would they result in different outputs ?

  ---

  Output from Python :

  ffmpeg version 4.1.4 Copyright (c) 2000-2019 the FFmpeg developers
  
    built with Apple LLVM version 10.0.1 (clang-1001.0.46.4)
  
    configuration: --prefix=/usr/local/Cellar/ffmpeg/4.1.4 --enable-shared --enable-pthreads --enable-version3 --enable-avresample --cc=clang --host-cflags='-I/Library/Java/JavaVirtualMachines/adoptopenjdk-12.0.1.jdk/Contents/Home/include -I/Library/Java/JavaVirtualMachines/adoptopenjdk-12.0.1.jdk/Contents/Home/include/darwin' --host-ldflags= --enable-ffplay --enable-gnutls --enable-gpl --enable-libaom --enable-libbluray --enable-libmp3lame --enable-libopus --enable-librubberband --enable-libsnappy --enable-libtesseract --enable-libtheora --enable-libvorbis --enable-libvpx --enable-libx264 --enable-libx265 --enable-libxvid --enable-lzma --enable-libfontconfig --enable-libfreetype --enable-frei0r --enable-libass --enable-libopencore-amrnb --enable-libopencore-amrwb --enable-libopenjpeg --enable-librtmp --enable-libspeex --enable-videotoolbox --disable-libjack --disable-indev=jack --enable-libaom --enable-libsoxr
  
    libavutil      56. 22.100 / 56. 22.100
  
    libavcodec     58. 35.100 / 58. 35.100
  
    libavformat    58. 20.100 / 58. 20.100
  
    libavdevice    58.  5.100 / 58.  5.100
  
    libavfilter     7. 40.101 /  7. 40.101
  
    libavresample   4.  0.  0 /  4.  0.  0
  
    libswscale      5.  3.100 /  5.  3.100
  
    libswresample   3.  3.100 /  3.  3.100
  
    libpostproc    55.  3.100 / 55.  3.100
  
  [mp3 @ 0x7ffa34001800] Estimating duration from bitrate, this may be inaccurate
  
  Input #0, mp3, from '/.../114006.mp3':
  
    Duration: 00:00:09.53, start: 0.000000, bitrate: 64 kb/s
  
      Stream #0:0: Audio: mp3, 22050 Hz, stereo, fltp, 64 kb/s
  
  Stream mapping:
  
    Stream #0:0 -> #0:0 (mp3 (mp3float) -> pcm_s16le (native))
  
  Press [q] to stop, [?] for help
  
  Output #0, wav, to '/.../114-006-a4dec52a.wav':
  
    Metadata:
  
      ISFT            : Lavf58.20.100
  
      Stream #0:0: Audio: pcm_s16le ([1][0][0][0] / 0x0001), 16000 Hz, mono, s16, 256 kb/s
  
      Metadata:
  
        encoder         : Lavc58.35.100 pcm_s16le
  
  size=     298kB time=00:00:09.53 bitrate= 256.1kbits/s speed= 166x    
  
  video:0kB audio:298kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: 0.025564%

  Output from terminal :

  ffmpeg version 4.1.4 Copyright (c) 2000-2019 the FFmpeg developers
  
    built with Apple LLVM version 10.0.1 (clang-1001.0.46.4)
  
    configuration: --prefix=/usr/local/Cellar/ffmpeg/4.1.4 --enable-shared --enable-pthreads --enable-version3 --enable-avresample --cc=clang --host-cflags='-I/Library/Java/JavaVirtualMachines/adoptopenjdk-12.0.1.jdk/Contents/Home/include -I/Library/Java/JavaVirtualMachines/adoptopenjdk-12.0.1.jdk/Contents/Home/include/darwin' --host-ldflags= --enable-ffplay --enable-gnutls --enable-gpl --enable-libaom --enable-libbluray --enable-libmp3lame --enable-libopus --enable-librubberband --enable-libsnappy --enable-libtesseract --enable-libtheora --enable-libvorbis --enable-libvpx --enable-libx264 --enable-libx265 --enable-libxvid --enable-lzma --enable-libfontconfig --enable-libfreetype --enable-frei0r --enable-libass --enable-libopencore-amrnb --enable-libopencore-amrwb --enable-libopenjpeg --enable-librtmp --enable-libspeex --enable-videotoolbox --disable-libjack --disable-indev=jack --enable-libaom --enable-libsoxr
  
    libavutil      56. 22.100 / 56. 22.100
  
    libavcodec     58. 35.100 / 58. 35.100
  
    libavformat    58. 20.100 / 58. 20.100
  
    libavdevice    58.  5.100 / 58.  5.100
  
    libavfilter     7. 40.101 /  7. 40.101
  
    libavresample   4.  0.  0 /  4.  0.  0
  
    libswscale      5.  3.100 /  5.  3.100
  
    libswresample   3.  3.100 /  3.  3.100
  
    libpostproc    55.  3.100 / 55.  3.100
  
  [mp3 @ 0x7f98ed005800] Estimating duration from bitrate, this may be inaccurate
  
  Input #0, mp3, from '/.../114006.mp3':
  
    Duration: 00:00:09.53, start: 0.000000, bitrate: 64 kb/s
  
      Stream #0:0: Audio: mp3, 22050 Hz, stereo, fltp, 64 kb/s
  
  Stream mapping:
  
    Stream #0:0 -> #0:0 (mp3 (mp3float) -> pcm_s16le (native))
  
  Press [q] to stop, [?] for help
  
  Output #0, wav, to '/.../114-006-a4dec52a.wav':
  
    Metadata:
  
      ISFT            : Lavf58.20.100
  
      Stream #0:0: Audio: pcm_s16le ([1][0][0][0] / 0x0001), 16000 Hz, mono, s16, 256 kb/s
  
      Metadata:
  
        encoder         : Lavc58.35.100 pcm_s16le
  
  size=     298kB time=00:00:09.53 bitrate= 256.1kbits/s speed= 469x    
  
  video:0kB audio:298kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: 0.025564%

• Failed to add audio to mp4 with moviepy

  24 avril, par D G

  I generated an audio sound.wav and a video temp.mp4 that makes use of the audio. Both with the same duration.

  


  I got the following warning when generating the temp.mp4. The animation is out of sync, it means that it freezes before the audio finishes.

  


  


  .venv\Lib\site-packages\moviepy\video\io\ffmpeg_reader.py:178 : UserWarning : In file temp.mp4, 1080000 bytes wanted but 0
bytes read at frame index 299 (out of a total 300 frames), at time 4.98/5.00 sec. Using the last valid frame instead.
warnings.warn(

  


  


  Complete code :

  


  # to generate sound.wav
import numpy as np
import soundfile as sf
from tqdm import tqdm
from os import startfile

# Parameters
filename = "sound.wav"
duration = 5  # seconds
num_voices = 1000
sample_rate = 44100  # Hz
chunk_size = sample_rate  # write in 1-second chunks


t = np.linspace(0, duration, int(sample_rate * duration), endpoint=False)

# Create many detuned sine waves (Deep Note style)
start_freqs = np.random.uniform(100, 400, num_voices)
end_freqs = np.linspace(400, 800, num_voices)  # target harmony

# Each voice sweeps from start to end frequency
signal = np.zeros_like(t)
for i in range(num_voices):
    freqs = np.linspace(start_freqs[i], end_freqs[i], t.size)
    voice = np.sin(2 * np.pi * freqs * t)
    voice *= np.sin(np.pi * i / num_voices)  # slight variance
    signal += voice

# Volume envelope
envelope = np.linspace(0.01, 1.0, t.size)
signal *= envelope

# Normalize
signal /= np.max(np.abs(signal))

# Save with progress bar using soundfile
with sf.SoundFile(
    filename, "w", samplerate=sample_rate, channels=1, subtype="PCM_16"
) as f:
    for i in tqdm(range(0, len(signal), chunk_size), desc=f"Saving {filename}"):
        f.write(signal[i : i + chunk_size])


startfile(filename)


  


  # to generate temp.mp4
from numpy import pi, sin, cos
import matplotlib.pyplot as plt
from matplotlib.animation import FuncAnimation
from os import startfile
from tqdm import tqdm
from moviepy import VideoFileClip, AudioFileClip, CompositeAudioClip

# Output settings
filename = "temp.mp4"
duration = 5  # seconds of animation
maxdim = 4  # canvas size in plot units (scaled)
fps = 60

# Real-world parameters
r = 40  # km (radius)
endtime = 2  # hours (duration of real motion)
rph = 0.5  # rotations per hour
omega = 2 * pi * rph  # rad/hour
speed = omega * r  # km/hour

# Animation setup
frames = duration * fps
scale = maxdim / r  # scale from km to plot units
dt = endtime / frames  # time per frame in hours

# Prepare figure and axes
fig, ax = plt.subplots(figsize=(6, 6))
ax.set_xlim(-maxdim - 1, maxdim + 1)
ax.set_ylim(-maxdim - 1, maxdim + 1)
ax.set_aspect("equal")
ax.grid()



# Plot circle path
circle = plt.Circle((0, 0), r * scale, color="lightgray", fill=False, linestyle="--")
ax.add_patch(circle)

# Moving point
(point,) = ax.plot([], [], "ro")

# Info text at center of the circle
info_text = ax.text(
    0, 0, "", fontsize=10,
    ha="center", va="center",
    bbox=dict(boxstyle="round,pad=0.4", facecolor="white", alpha=0.8)
)


def init():
    point.set_data([], [])
    info_text.set_text("")
    return point, info_text


def update(frame):
    t = frame * dt  # time in hours
    theta = omega * t  # angle in radians

    x = r * cos(theta) * scale
    y = r * sin(theta) * scale

    point.set_data([x], [y])
    info_text.set_text(
        f"Time: {t:.2f} hr\nRadius: {r:.1f} km\nSpeed: {speed:.2f} km/h"
    )
    return point, info_text



# Create animation
anim = FuncAnimation(
    fig, update, frames=frames, init_func=init, blit=True, interval=1000 / fps
)


with tqdm(total=frames, desc="Saving", unit="frame") as pbar:
    anim.save(filename, fps=fps, progress_callback=lambda i, n: pbar.update(1))

# Add sound using MoviePy
video = VideoFileClip(filename)
video.audio = CompositeAudioClip([AudioFileClip("sound.wav")])
video.write_videofile(filename)

startfile(filename)


  


  Could you figure out what the culprit is and how to fix it ?

  


  

  


  Edit

  


  Based on the given comment, I did the following but the problem still exists.

  


  # Add sound using MoviePy
video = VideoFileClip(filename)
audio = AudioFileClip("sound.wav")
audio.duration = video.duration
video.audio = CompositeAudioClip([audio])
video.write_videofile(filename)