Recherche avancée

Sur d’autres sites (11857)

• Empty output when extracting frames with FFMPEG at specific timestamps taken from FFPROBE

  9 octobre 2023, par Andrew McDowell

  I am trying to extract frames from a video as close as possible to specific timestamps, but for some timestamps I'm getting no output.

  


  I'm first using ffprobe to get a list of all frames along with their time values :

  


  ffprobe -select_streams v -show_frames -show_entries frame=pts_time,pkt_pts_time,pict_type <input video="video" />

  


  Which gives me an output like :

  


  [FRAME]
pts_time=11.950000
pict_type=I
[/FRAME]
[FRAME]
pts_time=11.966667
pict_type=B
[/FRAME]
[FRAME]
pts_time=11.983333
pict_type=P
[/FRAME]


  


  I'm then parsing this and determining which frame is the closest to the time I want and using FFmpeg to extract frames which match the chosen times like so :

  


  ffmpeg -i <input video="video" /> -filter:v "select=&#x27;eq(t, 11.950000)&#x2B;eq(t, 11.983333)&#x27;" -vsync drop -start_number 0 <output path="path">/frame-%08d.png</output>

  


  This mostly works but I've noticed for some frames I get no output. From some testing it always seems to work for I frames and works occasionally for P or B frames but not always.

  


  I can't see any pattern to which frames it works for or not and would love an insight into whether I'm doing something wrong in the extraction, or whether there's a way to determine which frames can be extracted or not.

  


• Python Av encode result is empty always

  19 septembre 2023, par Emir evcil

  I'm quite a newbie when it comes to ffmpeg and av, I was trying to do a simple encode - decode example but the encode result always returns an empty list. I don't understand the reason for this, I don't know how it can be solved.

  


  import av
import numpy as np
import cv2 

container = av.open(file = "video=PC Camera",format = "dshow",options = {"video_size":"640x480"},mode = "r")

def getFrame():
    # open camera
    for frame in container.decode(video=0):
        yield frame
        
def encode(frame):
    encoder = av.CodecContext.create('h264', 'w')
    encoder.width = frame.width
    encoder.height = frame.height
    encoder.pix_fmt = frame.format.name
    encoder.bit_rate = 5000000
    encoder.framerate = 30
    encoder.open()
    frames = encoder.encode(frame)
    encoder.close()
    return frames
    
    
if __name__ == "__main__":
    while 1:
        frame = next(getFrame())
        frame_encoded = encode(frame)
        print(frame_encoded) // always empty
        if len(frame_encoded) == 0:
            continue
        #cv2.imshow("frame",frame.to_ndarray(format = "bgr24"))
        #cv2.waitKey(1)


  


• Empty error object produced by ffprobe in Google Cloud Function

  20 septembre 2023, par willbattel

  Update : After more digging I found an open GitHub issue where others appear to be encountering the same behavior.

  


  ---

  


  I have a Google Cloud Function (2nd gen) in which I am trying to use ffprobe to get metadata from a video file stored in a Google Cloud Storage bucket. It is my understanding that I can generate a signed url and, by passing that directly to ffprobe, avoid loading the entire video file into memory. I generate a signed url and pass it to ffprobe, and then parse the output like so :

  


  import ffmpeg from 'fluent-ffmpeg'
import ffprobeStatic from 'ffprobe-static'

async function getVideoData(srcFile: File) {
    const [signedUrl] = await srcFile.getSignedUrl({
        action: 'read',
        expires: (new Date()).getMilliseconds() + 60_000,
    })

    const videoData: ffmpeg.FfprobeData = await new Promise((resolve, reject) => {
        ffmpeg.setFfprobePath(ffprobeStatic.path)
        ffmpeg.ffprobe(signedUrl, (err, data) => {
            if (err) {
                reject(err)
            }
            else {
                resolve(data)
            }
        })
    })

    return videoData
}


  


  This code works (with the same signed URL) locally on my macOS machine, but does not when deployed in a 2nd generation Google Cloud Function. In the latter case, data is undefined and err is {}.

  


  My main question is how to properly use ffprobe in 2nd gen Google Cloud Functions. I have tried to research this but documentation on ffmpeg/ffprobe on GCP is sparse. I'm also trying to figure out exactly why the error object err is empty...it's not very helpful 😅

  


  Additional info :

  


  

  • Environment : Google Cloud Functions 2nd Gen

  


  • Runtime : Node 20

  


  • "ffprobe-static" : "3.1.0",

  


  • "fluent-ffmpeg" : "2.1.2"

  


  


  Thanks in advance.