Recherche avancée

Sur d’autres sites (8632)

• Accented characters are not recognized in python [closed]

  10 avril 2023, par CorAnna

  I have a problem in the python script, my script should put subtitles in a video given a srt file, this srt file is written by another script but in its script it replaces the accents and all the particular characters with a black square symbol with a question mark inside it... the problem I think lies in the writing of this file, what follows and that in overwriting the subtitles I do with ffmpeg the sentences that contain an accented word are not written

  


  def video_audio_file_writer(video_file):

    videos_folder = "Video"
    audios_folder = "Audio"

    video_path = f"{videos_folder}\\{video_file}"

    video_name = Path(video_path).stem
    audio_name = f"{video_name}Audio"

    audio_path = f"{audios_folder}\\{audio_name}.wav"

    video = mp.VideoFileClip(video_path)
    audio = video.audio.write_audiofile(audio_path)

    return video_path, audio_path, video_name

    def audio_file_transcription(audio_path, lang):

    model = whisper.load_model("base")
    tran = gt.Translator()

    audio_file = str(audio_path)

    options = dict(beam_size=5, best_of=5)
    translate = dict(task="translate", **options)
    result = model.transcribe(audio_file, **translate)   

    return result

def audio_subtitles_transcription(result, video_name):

    subtitle_folder = "Content"
    subtitle_name = f"{video_name}Subtitle"
    subtitle_path_form = "srt"

    subtitle_path = f"{subtitle_folder}\\{subtitle_name}.{subtitle_path_form}"

    with open(os.path.join(subtitle_path), "w") as srt:
        # write_vtt(result["segments"], file=vtt)
        write_srt(result["segments"], file=srt)
            
    return subtitle_path

def video_subtitles(video_path, subtitle_path, video_name):

    video_subtitled_folder = "VideoSubtitles"
    video_subtitled_name = f"{video_name}Subtitles"
    video_subtitled_path = f"{video_subtitled_folder}\\{video_subtitled_name}.mp4"

    video_path_b = bytes(video_path, 'utf-8')
    subtitle_path_b = bytes(subtitle_path, 'utf-8')
    video_subtitled_path_b = bytes(video_subtitled_path, 'utf-8')

    path_abs_b = os.getcwdb() + b"\\"

    path_abs_bd = path_abs_b.decode('utf-8')
    video_path_bd= video_path_b.decode('utf-8')
    subtitle_path_bd = subtitle_path_b.decode('utf-8')
    video_subtitled_path_bd = video_subtitled_path_b.decode('utf-8')

    video_path_abs = str(path_abs_bd + video_path_bd)
    subtitle_path_abs = str(path_abs_bd + subtitle_path_bd).replace("\\", "\\\\").replace(":", "\\:")
    video_subtitled_path_abs = str(path_abs_bd + video_subtitled_path_bd)

    time.sleep(3)

    os.system(f"ffmpeg -i {video_path_abs} -vf subtitles='{subtitle_path_abs}' -y {video_subtitled_path_abs}")

    return video_subtitled_path_abs, video_path_abs, subtitle_path_abs

if __name__ == "__main__":

    video_path, audio_path, video_name = video_audio_file_writer(video_file="ChiIng.mp4")
    result = audio_file_transcription(audio_path=audio_path, lang="it")
    subtitle_path = audio_subtitles_transcription(result=result, video_name=video_name)
    video_subtitled_path_abs, video_path_abs, subtitle_path_abs = video_subtitles(video_path=video_path, subtitle_path=subtitle_path, video_name=video_name)
    
    print("Video Subtitled")


  


  Windows 11
Python 3.10

  


• Bit-field badness

  30 janvier 2010, par Mans — Compilers, Optimisation

  Consider the following C code which is based on an real-world situation.

  struct bf1_31 
      unsigned a:1 ;
      unsigned b:31 ;
   ;
  void func(struct bf1_31 *p, int n, int a)
  
  
  
      int i = 0 ;
  
      do 
  
          if (p[i].a)
  
              p[i].b += a ;
  
       while (++i < n) ;
  
  
  

  How would we best write this in ARM assembler ? This is how I would do it :
  

  func :
          ldr     r3,  [r0], #4
          tst     r3,  #1
          add     r3,  r3,  r2,  lsl #1
          strne   r3,  [r0, #-4]
          subs    r1,  r1,  #1
          bgt     func
          bx      lr
  

  The add instruction is unconditional to avoid a dependency on the comparison. Unrolling the loop would mask the latency of the ldr instruction as well, but that is outside the scope of this experiment.

  Now compile this code with gcc -march=armv5te -O3 and watch in horror :

  func :
          push    r4
          mov     ip, #0
          mov     r4, r2
  loop :
          ldrb    r3, [r0]
          add     ip, ip, #1
          tst     r3, #1
          ldrne   r3, [r0]
          andne   r2, r3, #1
          addne   r3, r4, r3, lsr #1
          orrne   r2, r2, r3, lsl #1
          strne   r2, [r0]
          cmp     ip, r1
          add     r0, r0, #4
          blt     loop
          pop     r4
          bx      lr
  

  This is nothing short of awful :

  • The same value is loaded from memory twice.
  • A complicated mask/shift/or operation is used where a simple shifted add would suffice.
  • Write-back addressing is not used.
  • The loop control counts up and compares instead of counting down.
  • Useless mov in the prologue ; swapping the roles or r2 and r4 would avoid this.
  • Using lr in place of r4 would allow the return to be done with pop {pc}, saving one instruction (ignoring for the moment that no callee-saved registers are needed at all).

  Even for this trivial function the gcc-generated code is more than twice the optimal size and slower by approximately the same factor.

  The main issue I wanted to illustrate is the poor handling of bit-fields by gcc. When accessing bitfields from memory, gcc issues a separate load for each field even when they are contained in the same aligned memory word. Although each load after the first will most likely hit L1 cache, this is still bad for several reasons :

  • Loads have typically two or three cycles result latency compared to one cycle for data processing instructions. Any bit-field can be extracted from a register with two shifts, and on ARM the second of these can generally be achieved using a shifted second operand to a following instruction. The ARMv6T2 instruction set also adds the SBFX and UBFX instructions for extracting any signed or unsigned bit-field in one cycle.
  • Most CPUs have more data processing units than load/store units. It is thus more likely for an ALU instruction than a load/store to issue without delay on a superscalar processor.
  • Redundant memory accesses can trigger early flushing of store buffers rendering these less efficient.

  No gcc bashing is complete without a comparison with another compiler, so without further ado, here is the ARM RVCT output (armcc --cpu 5te -O3) :

  func :
          mov     r3, #0
          push    r4, lr
  loop :
          ldr     ip, [r0, r3, lsl #2]
          tst     ip, #1
          addne   ip, ip, r2, lsl #1
          strne   ip, [r0, r3, lsl #2]
          add     r3, r3, #1
          cmp     r3, r1
          blt     loop
          pop     r4, pc
  

  This is much better, the core loop using only one instruction more than my version. The loop control is counting up, but at least this register is reused as offset for the memory accesses. More remarkable is the push/pop of two registers that are never used. I had not expected to see this from RVCT.

  Even the best compilers are still no match for a human.

• Batch splitting large audio files into small fixed-length audio files in moments of silence

  26 juillet 2023, par Haldjärvi

  to train the SO-VITS-SVC neural network, we need 10-14 second voice files. As a material, let's say I use phrases from some game. I have already made a batch script for decoding different files into one working format, another batch script for removing silence, as well as a batch script for combining small audio files into files of 13-14 seconds (I used Python, pydub and FFmpeg). To successfully automatically create a training dataset, it remains only to make one batch script - Cutting audio files lasting more than 14 seconds into separate files lasting 10-14 seconds, cutting in places of silence or close to silence is highly preferable.

  


  So, it is necessary to batch cut large audio files (20 seconds, 70 seconds, possibly several hundred seconds) into segments of approximately 10-14 seconds, however, the main task is to look for the quietest place in the cut areas so as not to cut phrases in the middle of a word (this is not very good for model training). So, is it really possible to do this in a very optimal way, so that the processing of a 30-second file does not take 15 seconds, but is fast ? Quiet zone detection is required only in the area of cuts, that is, 10-14 seconds, if counted from the very beginning of the file.

  


  I would be very grateful for any help.

  


  I tried to write a script together with ChatGPT, but all options gave completely unpredictable results and were not even close to what I needed... I had to stop at the option with a sharp cut of files for exactly 14000 milliseconds. However, I hope there is a chance to make a variant with cutting exactly in quiet areas.

  


  import os
from pydub import AudioSegment

input_directory = ".../RemSilence/"
output_directory = ".../Split/"
max_duration = 14000

def split_audio_by_duration(input_file, duration):
    audio = AudioSegment.from_file(input_file)
    segments = []
    for i in range(0, len(audio), duration):
        segment = audio[i:i + duration]
        segments.append(segment)
    return segments

if __name__ == "__main__":
    os.makedirs(output_directory, exist_ok=True)
    audio_files = [os.path.join(input_directory, file) for file in os.listdir(input_directory) if file.endswith(".wav")]
    audio_files.sort(key=lambda file: len(AudioSegment.from_file(file)))
    for file in audio_files:
        audio = AudioSegment.from_file(file)
        if len(audio) > max_duration:
            segments = split_audio_by_duration(file, max_duration)
            for i, segment in enumerate(segments):
                output_filename = f"output_{len(os.listdir(output_directory))+1}.wav"
                output_file_path = os.path.join(output_directory, output_filename)
                segment.export(output_file_path, format="wav")
        else:
            output_filename = f"output_{len(os.listdir(output_directory))+1}.wav"
            output_file_path = os.path.join(output_directory, output_filename)
            audio.export(output_file_path, format="wav")