Recherche avancée

Sur d’autres sites (13600)

• Python asyncio subprocess code returns "pipe closed by peer or os.write(pipe, data) raised exception."

  4 novembre 2022, par Duke Dougal

  I am trying to convert a synchronous Python process to asyncio. Any ideas what I am doing wrong ?

  


  This is the synchronous code which successfully starts ffmpeg and converts a directory of webp files into a video.

  


  import subprocess
import shlex
from os import listdir
from os.path import isfile, join

output_filename = 'output.mp4'
process = subprocess.Popen(shlex.split(f'ffmpeg -y -framerate 60 -i pipe: -vcodec libx265 -pix_fmt yuv420p -crf 24 output.mp4'), stdin=subprocess.PIPE)

thepath = '/home/ubuntu/webpfiles/'
thefiles = [f for f in listdir(thepath) if isfile(join(thepath, f))]
for filename in thefiles:
    absolute_path = f'{thepath}{filename}'
    with open(absolute_path, 'rb') as f:
        process.stdin.write(f.read())

process.stdin.close()
process.wait()
process.terminate()


  


  This async code fails :

  


  from os import listdir
from os.path import isfile, join
import shlex
import asyncio

outputfilename = 'output.mp4'

async def write_stdin(proc):
    thepath = '/home/ubuntu/webpfiles/'
    thefiles = [f for f in listdir(thepath) if isfile(join(thepath, f))]
    thefiles.sort()
    for filename in thefiles:
        absolute_path = f'{thepath}{filename}'
        with open(absolute_path, 'rb') as f:
            await proc.communicate(input=f.read())

async def create_ffmpeg_subprocess():
    bin = f'/home/ubuntu/bin/ffmpeg'
    params = f'-y -framerate 60 -i pipe: -vcodec libx265 -pix_fmt yuv420p -crf 24 {outputfilename}'
    proc = await asyncio.create_subprocess_exec(
        bin,
        *shlex.split(params),
        stdin=asyncio.subprocess.PIPE,
        stdout=asyncio.subprocess.PIPE,
        stderr=asyncio.subprocess.PIPE,
    )
    return proc

async def start():
    loop = asyncio.get_event_loop()
    proc = await create_ffmpeg_subprocess()
    task_stdout = loop.create_task(write_stdin(proc))
    await asyncio.gather(task_stdout)

if __name__ == '__main__':
    asyncio.run(start())


  


  The output for the async code is :

  


  pipe closed by peer or os.write(pipe, data) raised exception.
pipe closed by peer or os.write(pipe, data) raised exception.
pipe closed by peer or os.write(pipe, data) raised exception.
pipe closed by peer or os.write(pipe, data) raised exception.
pipe closed by peer or os.write(pipe, data) raised exception.
pipe closed by peer or os.write(pipe, data) raised exception.
pipe closed by peer or os.write(pipe, data) raised exception.
pipe closed by peer or os.write(pipe, data) raised exception.
pipe closed by peer or os.write(pipe, data) raised exception.
pipe closed by peer or os.write(pipe, data) raised exception.
pipe closed by peer or os.write(pipe, data) raised exception.
pipe closed by peer or os.write(pipe, data) raised exception.
pipe closed by peer or os.write(pipe, data) raised exception.
pipe closed by peer or os.write(pipe, data) raised exception.
pipe closed by peer or os.write(pipe, data) raised exception.
pipe closed by peer or os.write(pipe, data) raised exception.


  


  etc - one line for each webp file

  


• What is the right way to put data to abuffersrc and pull from abufersink when I have multiple inputs with FFMPEG "afir" filter ?

  19 décembre 2022, par Evgenii

  I want to understand why av_buffersink_get_frame() always return AVERROR(EAGAIN).
I use this algorithm :

  


  

  1. Open input file input.mp3, create decoder for input file

  


  2. Open signal file signal.wav, create decoder for signal file

  


  3. Create filter graph and inputs/outputs with function avfilter_graph_parse2("[in0][in1]afir[out], &inputs, &outputs")

  


  4. Create abuffersrc for every inputs with names "in0", "in1"

  


  5. Create one abuffersink for one outputs with name "out"

  


  6. Read and decode samples with methods av_read_frame(), avcodec_send_packet(), avcodec_receive_frame(), put samples to abuffersrc with av_buffersrc_add_frame() and trying to read samples from abuffersink with av_buffersink_get_frame() - HERE I get AVERROR(EAGAIN) for every call

  


  


  Please, help !

  


  I've tried to read all samples and push its to abuffersrc for input and signal pipes in first step and and call av_buffersink_get_frame one time in second step. I got AVERROR(EGAIN) again.

  


  I've tried to config two outputs with names "in0" and "in1" and one input with name "out" and call avfilter_graph_parse_ptr() as done in here :

  


  static AVFilterContext *init_buffersrc(AVCodecContext *decoder, AVFilterGraph *filter_graph, const char *pad_name) {
    AVFilterContext *buffersrc = NULL;
    uint8_t args[1024];

    snprintf(args, sizeof(args),
        "time_base=%d/%d:sample_rate=%d:sample_fmt=%s:channel_layout=0x%"PRIx64"",
        decoder->time_base.num, decoder->time_base.den, decoder->sample_rate, av_get_sample_fmt_name(decoder->sample_fmt), decoder->channel_layout);

    CHECK_AV(avfilter_graph_create_filter(&buffersrc, avfilter_get_by_name("abuffer"), pad_name, args, NULL, filter_graph));
    return buffersrc;
}

static AVFilterContext *init_buffersink(AVFilterGraph *filter_graph) {
    AVFilterContext *buffersink = NULL;

    CHECK_AV(avfilter_graph_create_filter(&buffersink, avfilter_get_by_name("abuffersink"), "out", NULL, NULL, filter_graph));
    return buffersink;
}

static void init_complex_graph(const char *desc, Context *ctx) {

    AVFilterGraph *filter_graph;
    AVFilterInOut *inputs = avfilter_inout_alloc();
    AVFilterInOut **outputs = (AVFilterInOut**) av_malloc(ctx->nb_input_files*sizeof(AVFilterInOut *));
    char *dump;
    uint8_t pad_name[64];

    filter_graph = ctx->filter_graph = avfilter_graph_alloc();
    filter_graph->nb_threads = 1;

    for (int i = 0; i< ctx->nb_input_files; ++i) {
        outputs[i] = avfilter_inout_alloc();
    }

    for (int i = 0; i < ctx->nb_input_files; ++i) {
        AVFilterContext *buffersrc;
        snprintf(pad_name, sizeof(pad_name), "in%d", i);        
        buffersrc = init_buffersrc(ctx->input_files[i]->decoder, filter_graph, pad_name);
        ctx->input_files[i]->buffersrc = buffersrc;
        outputs[i]->name = av_strdup(pad_name);
        outputs[i]->filter_ctx = buffersrc;
        outputs[i]->pad_idx = 0;
        if (i == ctx->nb_input_files - 1) {
            outputs[i]->next = NULL;
        }
        else {
            outputs[i]->next = outputs[i + 1];
        }
    }

    ctx->output_file->buffersink = init_buffersink(filter_graph);

    inputs->name = av_strdup("out");
    inputs->filter_ctx = ctx->output_file->buffersink;
    inputs->pad_idx =0;
    inputs->next = NULL;

    CHECK_AV(avfilter_graph_parse_ptr(filter_graph, desc, &inputs, outputs, NULL));

    CHECK_AV(avfilter_graph_config(filter_graph, NULL));
}   

// Call init_complex_graph("[in0][in1]afir[out]", SOME_CONTEXT);


  


  and still nothing.

  


• Add a sample "CAMM" data to specific PTS in a MP4 file

  11 janvier 2023, par Cheloute

  For testing purpose, I need to add programatically some samples to an existing MP4 video file but I can't understand how.
My goal is to add some CAMM1 records (exposure data) associated to each PTS multiple of 33366, that's to say : PTS=0, PTS=33366, PTS=66732, ...

  


  I already have an existing camm track atom in my file, so the idea is to duplicate all the existing CAMM samples (CAMM2 and CAMM3 data), add my CAMM1 samples to the existing CAMM data, and write all the samples again at the end of the mdat atom (moov atom is located after mdat atom in my case). Next I'll update the CAMM track atom (stts, stsc, stsz and stco) to index all the samples correctly.

  


  But I have 1 question I can't understand. STSC and STCO are dealing with offsets and STTS with time and delta, how can I do to link my CAMM1 records to PTS multiple of 33366 ?

  


  Thanks