Recherche avancée

Sur d’autres sites (14252)

• How to seamlessly concatenate multiple Opus files together without popping sound ?

  16 février, par Gurdie Derilus

  I have a large PCM file that I've split into N chunks (N being the # of threads), and I encode them in parallel into Opus files with FFmpeg.

  


  Note : All PCM files are 16-bit Little Endian, 2 channels, 48000 sample rate.

  


  I then concatenate the Opus files using FFmpeg's demuxer, but I can hear an audible pop sound between each segment.

  


  Opening this sample file in Audacity reveals the issue :
Notice the introduced pops in opus

  


  I created a simple and short Golang project on Github with a sample PCM file for easy testing. Note, not production code, so obviously not following any best practices here.

  


  #1, I suspected the pops might've been introduced while parallel encoding each PCM file to Opus files. This, however, wasn't the case.
Concatted Opus files vs Separate Opus files image.

  


  #2, using the concat filter works, however it reencodes the files, which is not doable in my case as it's too slow (these files can & do reach up to an hour). I know Opus files are chainable, so I can't imagine why they don't work flawlessly.

  


  #3, I heard that Opus has a 20ms frame size, so I split the file against that frame size, but this made no difference.

  


  chunkSize := largePcmFileStat.Size() / int64(runtime.GOMAXPROCS(0))
chunkSize = int64(roundUpToNearestMultiple(float64(chunkSize), 4))


  


  The entire sample looks like this :

  


  package main

import (
    "context"
    "fmt"
    "io"
    "log"
    "os"
)

func main() {
    // Grab large PCM file
    largePcmFile, err := os.Open("files/full_raw.pcm")
    if err != nil {
        log.Fatalln(err)
    }

    // Split into 2 chunks
    ByteRate := 2
    SampleRate := 48000
    Channels := 2
    Seconds := 20
    chunkSize := Seconds * Channels * SampleRate * ByteRate

    file1, err := encodePcmToOpus(context.TODO(), io.LimitReader(largePcmFile, int64(chunkSize)))
    if err != nil {
        log.Fatalln(err)
    }

    file2, err := encodePcmToOpus(context.TODO(), io.LimitReader(largePcmFile, int64(chunkSize)))
    if err != nil {
        log.Fatalln(err)
    }

    fmt.Println("Check if these play with no defects:", file1)
    fmt.Println("file1:", file1)
    fmt.Println("file2:", file2)
    fmt.Println()

    concatFile, err := concatOpusFiles(context.TODO(), []string{file1, file2})
    if err != nil {
        log.Fatalln(err)
    }

    fmt.Println("concatted file:", concatFile.Name())
}


  


• Adding background audio in FFMPEG, but quieting it if there is sound in another channel

  9 septembre 2021, par Connor Bell

  I'm appending several videos together with FFMPEG. Some of these videos have accompanying audio, and some do not.

  


  I'd like to add music in the background, and have found out how to do so from here. However, I'd like the background music to be (let's say 80%) quieter if there is already audio in that video. Note that all videos have a null audio track, so just checking for the existence of an audio track isn't sufficient.

  


  My current process is :

  


  

  • Take source videos, add a null audio source and upscale (The null audio source is required for ffmpeg-concat to work due to a bug, I think)

  


  • Combine the videos using ffmpeg-concat

  


  


  Preferably, adding the background music should be the third step, as splitting the background music prior to combining the videos sounds more complex, but I may be wrong.

  


• Screeching white sound coming while playing audio as a raw stream

  27 avril 2020, par Sri Nithya Sharabheshwarananda

  I. Background

  



  

  1. I am trying to make an application which helps to match subtitles to the audio waveform very accurately at the waveform level, at the word level or even at the character level.

  


  2. The audio is expected to be Sanskrit chants (Yoga, rituals etc.) which are extremely long compound words [ example - aṅganyā-sokta-mātaro-bījam is traditionally one word broken only to assist reading ]

  


  3. The input transcripts / subtitles might be roughly in sync at the sentence/verse level but surely would not be in sync at the word level.

  


  4. The application should be able to figure out points of silence in the audio waveform, so that it can guess the start and end points of each word (or even letter/consonant/vowel in a word), such that the audio-chanting and visual-subtitle at the word level (or even at letter/consonant/vowel level) perfectly match, and the corresponding UI just highlights or animates the exact word (or even letter) in the subtitle line which is being chanted at that moment, and also show that word (or even the letter/consonant/vowel) in bigger font. This app's purpose is to assist learning Sanskrit chanting.

  


  5. It is not expected to be a 100% automated process, nor 100% manual but a mix where the application should assist the human as much as possible.

  


  



  II. Following is the first code I wrote for this purpose, wherein

  



  

  1. First I open a mp3 (or any audio format) file,

  


  2. Seek to some arbitrary point in the timeline of the audio file // as of now playing from zero offset

  


  3. Get the audio data in raw format for 2 purposes - (1) playing it and (2) drawing the waveform.

  


  4. Playing the raw audio data using standard java audio libraries

  


  



  III. The problem I am facing is, between every cycle there is screeching sound.

  



  

  • Probably I need to close the line between cycles ? Sounds simple, I can try.

  


  • But I am also wondering if this overall approach itself is correct ? Any tip, guide, suggestion, link would be really helpful.

  


  • Also I just hard coded the sample-rate etc ( 44100Hz etc. ), are these good to set as default presets or it should depend on the input format ?

  


  



  IV. Here is the code

  



  import com.github.kokorin.jaffree.StreamType;
import com.github.kokorin.jaffree.ffmpeg.FFmpeg;
import com.github.kokorin.jaffree.ffmpeg.FFmpegProgress;
import com.github.kokorin.jaffree.ffmpeg.FFmpegResult;
import com.github.kokorin.jaffree.ffmpeg.NullOutput;
import com.github.kokorin.jaffree.ffmpeg.PipeOutput;
import com.github.kokorin.jaffree.ffmpeg.ProgressListener;
import com.github.kokorin.jaffree.ffprobe.Stream;
import com.github.kokorin.jaffree.ffmpeg.UrlInput;
import com.github.kokorin.jaffree.ffprobe.FFprobe;
import com.github.kokorin.jaffree.ffprobe.FFprobeResult;
import java.io.IOException;
import java.io.OutputStream;
import java.nio.file.Path;
import java.nio.file.Paths;
import java.util.concurrent.TimeUnit;
import java.util.concurrent.atomic.AtomicLong;
import javax.sound.sampled.AudioFormat;
import javax.sound.sampled.AudioSystem;
import javax.sound.sampled.DataLine;
import javax.sound.sampled.SourceDataLine;


public class FFMpegToRaw {
    Path BIN = Paths.get("f:\\utilities\\ffmpeg-20190413-0ad0533-win64-static\\bin");
    String VIDEO_MP4 = "f:\\org\\TEMPLE\\DeviMahatmyamRecitationAudio\\03_01_Devi Kavacham.mp3";
    FFprobe ffprobe;
    FFmpeg ffmpeg;

    public void basicCheck() throws Exception {
        if (BIN != null) {
            ffprobe = FFprobe.atPath(BIN);
        } else {
            ffprobe = FFprobe.atPath();
        }
        FFprobeResult result = ffprobe
                .setShowStreams(true)
                .setInput(VIDEO_MP4)
                .execute();

        for (Stream stream : result.getStreams()) {
            System.out.println("Stream " + stream.getIndex()
                    + " type " + stream.getCodecType()
                    + " duration " + stream.getDuration(TimeUnit.SECONDS));
        }    
        if (BIN != null) {
            ffmpeg = FFmpeg.atPath(BIN);
        } else {
            ffmpeg = FFmpeg.atPath();
        }

        //Sometimes ffprobe can't show exact duration, use ffmpeg trancoding to NULL output to get it
        final AtomicLong durationMillis = new AtomicLong();
        FFmpegResult fFmpegResult = ffmpeg
                .addInput(
                        UrlInput.fromUrl(VIDEO_MP4)
                )
                .addOutput(new NullOutput())
                .setProgressListener(new ProgressListener() {
                    @Override
                    public void onProgress(FFmpegProgress progress) {
                        durationMillis.set(progress.getTimeMillis());
                    }
                })
                .execute();
        System.out.println("audio size - "+fFmpegResult.getAudioSize());
        System.out.println("Exact duration: " + durationMillis.get() + " milliseconds");
    }

    public void toRawAndPlay() throws Exception {
        ProgressListener listener = new ProgressListener() {
            @Override
            public void onProgress(FFmpegProgress progress) {
                System.out.println(progress.getFrame());
            }
        };

        // code derived from : https://stackoverflow.com/questions/32873596/play-raw-pcm-audio-received-in-udp-packets

        int sampleRate = 44100;//24000;//Hz
        int sampleSize = 16;//Bits
        int channels   = 1;
        boolean signed = true;
        boolean bigEnd = false;
        String format  = "s16be"; //"f32le"

        //https://trac.ffmpeg.org/wiki/audio types
        final AudioFormat af = new AudioFormat(sampleRate, sampleSize, channels, signed, bigEnd);
        final DataLine.Info info = new DataLine.Info(SourceDataLine.class, af);
        final SourceDataLine line = (SourceDataLine) AudioSystem.getLine(info);

        line.open(af, 4096); // format , buffer size
        line.start();

        OutputStream destination = new OutputStream() {
            @Override public void write(int b) throws IOException {
                throw new UnsupportedOperationException("Nobody uses thi.");
            }
            @Override public void write(byte[] b, int off, int len) throws IOException {
                String o = new String(b);
                boolean showString = false;
                System.out.println("New output ("+ len
                        + ", off="+off + ") -> "+(showString?o:"")); 
                // output wave form repeatedly

                if(len%2!=0) {
                    len -= 1;
                    System.out.println("");
                }
                line.write(b, off, len);
                System.out.println("done round");
            }
        };

        // src : http://blog.wudilabs.org/entry/c3d357ed/?lang=en-US
        FFmpegResult result = FFmpeg.atPath(BIN).
            addInput(UrlInput.fromPath(Paths.get(VIDEO_MP4))).
            addOutput(PipeOutput.pumpTo(destination).
                disableStream(StreamType.VIDEO). //.addArgument("-vn")
                setFrameRate(sampleRate).            //.addArguments("-ar", sampleRate)
                addArguments("-ac", "1").
                setFormat(format)              //.addArguments("-f", format)
            ).
            setProgressListener(listener).
            execute();

        // shut down audio
        line.drain();
        line.stop();
        line.close();

        System.out.println("result = "+result.toString());
    }

    public static void main(String[] args) throws Exception {
        FFMpegToRaw raw = new FFMpegToRaw();
        raw.basicCheck();
        raw.toRawAndPlay();
    }
}



  



  Thank You