Recherche avancée

Sur d’autres sites (10545)

• Start and end time of MoviePy's VideoClip not working

  21 mars 2024, par ernesto casco velazquez

  I'm trying to add captions to a video. The desired outcome is to show each word in the exact moment is being said.

  


  I have a method that gives me the accurate time start and end per each word :

  


  def get_words_per_time(audio_speech_file):
    model = whisper.load_model("base")
    transcribe = model.transcribe(
        audio=audio_speech_file, fp16=False, word_timestamps=True
    )
    segments = transcribe["segments"]
    words = []

    for seg in segments:
        for word in seg["words"]:
            words.append(
                {
                    "word": word["word"],
                    "start": word["start"],
                    "end": word["end"],
                    "prob": round(word["probability"], 4),
                }
            )
    return words


  


  Then I have a code that uses MoviePy to create TextClip and assing a given start and end time per pair of words (I know there are redundant statements, srry) :

  


  def generate_captions(
    words,
    font="Komika",
    fontsize=32,
    color="White",
    align="center",
    stroke_width=3,
    stroke_color="black",
):
    text_comp = []
    for i in track(range(0, len(words), 2), description="Creating captions..."):
        word1 = words[i]
        if i + 1 < len(words):
            word2 = words[i + 1]
        text_clip = TextClip(
            f"{word1['word']} {word2['word'] if i + 1 < len(words) else ''}",
            font=font,  # Change Font if not found
            fontsize=fontsize,
            color=color,
            align=align,
            method="caption",
            size=(660, None),
            stroke_width=stroke_width,
            stroke_color=stroke_color,
        )
        text_clip = text_clip.set_start(word1["start"])
        text_clip = text_clip.set_end(
            word2["end"] if i + 1 < len(words) else word1["end"]
        )
        text_comp.append(text_clip)
    return text_comp


  


  Finally, I concatenate the words into a single video :

  


  vid_clip = CompositeVideoClip(
    [vid_clip, concatenate_videoclips(text_comp).set_position(("center", 860))]
)


  


  The output is this, but you can clearly see the words are not flowing with the speech. They somehow move faster as if the start/end time did not matter. Here's the video

  


  The words with their respective start/end time, look like this :

  


  [
    {
        'word': 'This',
        'start': 0.0,
        'end': 0.22,
        'prob': 0.805
    },
    {
        'word': 'is',
        'start': 0.22,
        'end': 0.42,
        'prob': 0.9991
    },
    {
        'word': 'a',
        'start': 0.42,
        'end': 0.6,
        'prob': 0.999
    },
    {
        'word': 'test,
        ',
        'start': 0.6,
        'end': 1.04,
        'prob': 0.9939
    },
    {
        'word': 'to',
        'start': 1.18,
        'end': 1.3,
        'prob': 0.9847
    },
    {
        'word': 'show',
        'start': 1.3,
        'end': 1.54,
        'prob': 0.9971
    },
    {
        'word': 'words',
        'start': 1.54,
        'end': 1.9,
        'prob': 0.995
    },
    {
        'word': 'does',
        'start': 1.9,
        'end': 2.16,
        'prob': 0.997
    },
    {
        'word': 'not',
        'start': 2.16,
        'end': 2.4,
        'prob': 0.9978
    },
    {
        'word': 'appear.',
        'start': 2.4,
        'end': 2.82,
        'prob': 0.9984
    },
    {
        'word': 'At',
        'start': 3.46,
        'end': 3.6,
        'prob': 0.9793
    },
    {
        'word': 'their',
        'start': 3.6,
        'end': 3.8,
        'prob': 0.9984
    },
    {
        'word': 'proper',
        'start': 3.8,
        'end': 4.22,
        'prob': 0.9976
    },
    {
        'word': 'time.',
        'start': 4.22,
        'end': 4.72,
        'prob': 0.999
    },
    {
        'word': 'Thanks',
        'start': 5.04,
        'end': 5.4,
        'prob': 0.9662
    },
    {
        'word': 'for,
        ',
        'start': 5.4,
        'end': 5.66,
        'prob': 0.9941
    },
    {
        'word': 'watching.',
        'start': 5.94,
        'end': 6.36,
        'prob': 0.7701
    }
]


  


  What could be causing this ?

  


• Concatenate/Synchronize timestamp and video FFMPEG raspberry

  23 janvier 2019, par Andrea

  I have a raspberry pi 3 B+ and a logitech c920 webcam. I need to stream the webcam source from the raspberry along with the timestamp (per each frame) by maintaining 30fps.

  I used the following command :

  ffmpeg -ar 44100 -ac 2 -f alsa -i hw:1,0 -f v4l2 -codec:v h264 -framerate 30 -video_size 1920x1080 -itsoffset 0.5 -vsync 0 -copyts -i /dev/video0 -copyinkf -codec:v copy -use_wallclock_as_timestamps 1 -flags +global_header -codec:a aac -ab 128k -g 10 -vsync 0 -f mpegts udp://ip_of_my_machine:1234 -vsync 0 -f mkvtimestamp_v2 udp://ip_of_my_machine:1235

  I generate 2 udp streams but if I get the outpt with mplayer and netcat, respectively, I see a displacement between the timestamp in mplayer and the timestamp streamed of about 2 seconds.

  I need to get the two information separated because I have to process them and to maintain 30 FPS I cannot encode the output of the video.

  Is there a way to do it ?

• Record video using ffmpeg on mac os

  4 août 2015, par yogi

  I want to record live webcam video on mac OS using ffmpeg in any format. I tried a lot but not able to find out command for recoding of video. So Please any one can tell me ffmpeg command for capture video using web cam for mac os.

  Thanx in advance.