Recherche avancée

Sur d’autres sites (9140)

• FFmpeg RTSP drop rate increases when frame rate is reduced

  13 avril 2024, par Avishka Perera

  I need to read an RTSP stream, process the images individually in Python, and then write the images back to an RTSP stream. As the RTSP server, I am using Mediamtx [1]. For streaming, I am using FFmpeg [2].

  


  I have the following code that works perfectly fine. For simplification purposes, I am streaming three generated images.

  


  import time
import numpy as np
import subprocess

width, height = 640, 480
fps = 25
rtsp_server_address = f"rtsp://localhost:8554/mystream"

ffmpeg_cmd = [
    "ffmpeg",
    "-re",
    "-f",
    "rawvideo",
    "-pix_fmt",
    "rgb24",
    "-s",
    f"{width}x{height}",
    "-i",
    "-",
    "-r",
    str(fps),
    "-avoid_negative_ts",
    "make_zero",
    "-vcodec",
    "libx264",
    "-threads",
    "4",
    "-f",
    "rtsp",
    rtsp_server_address,
]
colors = np.array(
    [
        [255, 0, 0],
        [0, 255, 0],
        [0, 0, 255],
    ]
).reshape(3, 1, 1, 3)
images = (np.ones((3, width, height, 3)) * colors).astype(np.uint8)

if __name__ == "__main__":

    process = subprocess.Popen(ffmpeg_cmd, stdin=subprocess.PIPE)
    start = time.time()
    exported = 0
    while True:
        exported += 1
        next_time = start + exported / fps
        now = time.time()
        if next_time > now:
            sleep_dur = next_time - now
            time.sleep(sleep_dur)

        image = images[exported % 3]
        image_bytes = image.tobytes()

        process.stdin.write(image_bytes)
        process.stdin.flush()

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


  


  The issue is, that I need to run this at 10 fps because the processing step is heavy and can only afford 10 fps. Hence, as I reduce the frame rate from 25 to 10, the drop rate increases from 0% to 100%. And after a few iterations, I get a BrokenPipeError: [Errno 32] Broken pipe. Refer to the appendix for the complete log.

  


  As an alternative, I can use OpenCV compiled from source with GStreamer [3], but I prefer using FFmpeg to make the shipping process simple. Since compiling OpenCV from source can be tedious and dependent on the system.

  


  References

  


  [1] Mediamtx (formerly rtsp-simple-server) : https://github.com/bluenviron/mediamtx

  


  [2] FFmpeg : https://github.com/FFmpeg/FFmpeg

  


  [3] Compile OpenCV with GStreamer : https://github.com/bluenviron/mediamtx?tab=readme-ov-file#opencv

  


  Appendix

  


  Creating the source stream

  


  To instantiate the unprocessed stream, I use the following command. This streams the content of my webcam as and RTSP stream.

  


  ffmpeg -video_size 1280x720 -i /dev/video0  -avoid_negative_ts make_zero -vcodec libx264 -r 10 -f rtsp rtsp://localhost:8554/webcam


  


  Error log

  


  ffmpeg version 6.1.1 Copyright (c) 2000-2023 the FFmpeg developers&#xA;  built with gcc 12.3.0 (conda-forge gcc 12.3.0-5)&#xA;  configuration: --prefix=/home/conda/feedstock_root/build_artifacts/ffmpeg_1712656518955/_h_env_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_plac --cc=/home/conda/feedstock_root/build_artifacts/ffmpeg_1712656518955/_build_env/bin/x86_64-conda-linux-gnu-cc --cxx=/home/conda/feedstock_root/build_artifacts/ffmpeg_1712656518955/_build_env/bin/x86_64-conda-linux-gnu-c&#x2B;&#x2B; --nm=/home/conda/feedstock_root/build_artifacts/ffmpeg_1712656518955/_build_env/bin/x86_64-conda-linux-gnu-nm --ar=/home/conda/feedstock_root/build_artifacts/ffmpeg_1712656518955/_build_env/bin/x86_64-conda-linux-gnu-ar --disable-doc --disable-openssl --enable-demuxer=dash --enable-hardcoded-tables --enable-libfreetype --enable-libharfbuzz --enable-libfontconfig --enable-libopenh264 --enable-libdav1d --enable-gnutls --enable-libmp3lame --enable-libvpx --enable-libass --enable-pthreads --enable-vaapi --enable-libopenvino --enable-gpl --enable-libx264 --enable-libx265 --enable-libaom --enable-libsvtav1 --enable-libxml2 --enable-pic --enable-shared --disable-static --enable-version3 --enable-zlib --enable-libopus --pkg-config=/home/conda/feedstock_root/build_artifacts/ffmpeg_1712656518955/_build_env/bin/pkg-config&#xA;  libavutil      58. 29.100 / 58. 29.100&#xA;  libavcodec     60. 31.102 / 60. 31.102&#xA;  libavformat    60. 16.100 / 60. 16.100&#xA;  libavdevice    60.  3.100 / 60.  3.100&#xA;  libavfilter     9. 12.100 /  9. 12.100&#xA;  libswscale      7.  5.100 /  7.  5.100&#xA;  libswresample   4. 12.100 /  4. 12.100&#xA;  libpostproc    57.  3.100 / 57.  3.100&#xA;Input #0, rawvideo, from &#x27;fd:&#x27;:&#xA;  Duration: N/A, start: 0.000000, bitrate: 184320 kb/s&#xA;  Stream #0:0: Video: rawvideo (RGB[24] / 0x18424752), rgb24, 640x480, 184320 kb/s, 25 tbr, 25 tbn&#xA;Stream mapping:&#xA;  Stream #0:0 -> #0:0 (rawvideo (native) -> h264 (libx264))&#xA;[libx264 @ 0x5e2ef8b01340] using cpu capabilities: MMX2 SSE2Fast SSSE3 SSE4.2 AVX FMA3 BMI2 AVX2&#xA;[libx264 @ 0x5e2ef8b01340] profile High 4:4:4 Predictive, level 2.2, 4:4:4, 8-bit&#xA;[libx264 @ 0x5e2ef8b01340] 264 - core 164 r3095 baee400 - H.264/MPEG-4 AVC codec - Copyleft 2003-2022 - http://www.videolan.org/x264.html - options: cabac=1 ref=3 deblock=1:0:0 analyse=0x3:0x113 me=hex subme=7 psy=1 psy_rd=1.00:0.00 mixed_ref=1 me_range=16 chroma_me=1 trellis=1 8x8dct=1 cqm=0 deadzone=21,11 fast_pskip=1 chroma_qp_offset=4 threads=4 lookahead_threads=1 sliced_threads=0 nr=0 decimate=1 interlaced=0 bluray_compat=0 constrained_intra=0 bframes=3 b_pyramid=2 b_adapt=1 b_bias=0 direct=1 weightb=1 open_gop=0 weightp=2 keyint=250 keyint_min=10 scenecut=40 intra_refresh=0 rc_lookahead=40 rc=crf mbtree=1 crf=23.0 qcomp=0.60 qpmin=0 qpmax=69 qpstep=4 ip_ratio=1.40 aq=1:1.00&#xA;Output #0, rtsp, to &#x27;rtsp://localhost:8554/mystream&#x27;:&#xA;  Metadata:&#xA;    encoder         : Lavf60.16.100&#xA;  Stream #0:0: Video: h264, yuv444p(tv, progressive), 640x480, q=2-31, 10 fps, 90k tbn&#xA;    Metadata:&#xA;      encoder         : Lavc60.31.102 libx264&#xA;    Side data:&#xA;      cpb: bitrate max/min/avg: 0/0/0 buffer size: 0 vbv_delay: N/A&#xA;[vost#0:0/libx264 @ 0x5e2ef8b01080] Error submitting a packet to the muxer: Broken pipe   &#xA;[out#0/rtsp @ 0x5e2ef8afd780] Error muxing a packet&#xA;[out#0/rtsp @ 0x5e2ef8afd780] video:1kB audio:0kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: unknown&#xA;frame=    1 fps=0.1 q=-1.0 Lsize=N/A time=00:00:04.70 bitrate=N/A dup=0 drop=70 speed=0.389x    &#xA;[libx264 @ 0x5e2ef8b01340] frame I:16    Avg QP: 6.00  size:   147&#xA;[libx264 @ 0x5e2ef8b01340] frame P:17    Avg QP: 9.94  size:   101&#xA;[libx264 @ 0x5e2ef8b01340] frame B:17    Avg QP: 9.94  size:    64&#xA;[libx264 @ 0x5e2ef8b01340] consecutive B-frames: 50.0%  0.0% 42.0%  8.0%&#xA;[libx264 @ 0x5e2ef8b01340] mb I  I16..4: 81.3% 18.7%  0.0%&#xA;[libx264 @ 0x5e2ef8b01340] mb P  I16..4: 52.9%  0.0%  0.0%  P16..4:  0.0%  0.0%  0.0%  0.0%  0.0%    skip:47.1%&#xA;[libx264 @ 0x5e2ef8b01340] mb B  I16..4:  0.0%  5.9%  0.0%  B16..8:  0.1%  0.0%  0.0%  direct: 0.0%  skip:94.0%  L0:56.2% L1:43.8% BI: 0.0%&#xA;[libx264 @ 0x5e2ef8b01340] 8x8 transform intra:15.4% inter:100.0%&#xA;[libx264 @ 0x5e2ef8b01340] coded y,u,v intra: 0.0% 0.0% 0.0% inter: 0.0% 0.0% 0.0%&#xA;[libx264 @ 0x5e2ef8b01340] i16 v,h,dc,p: 97%  0%  3%  0%&#xA;[libx264 @ 0x5e2ef8b01340] i8 v,h,dc,ddl,ddr,vr,hd,vl,hu:  0%  0% 100%  0%  0%  0%  0%  0%  0%&#xA;[libx264 @ 0x5e2ef8b01340] Weighted P-Frames: Y:52.9% UV:52.9%&#xA;[libx264 @ 0x5e2ef8b01340] ref P L0: 88.9%  0.0%  0.0% 11.1%&#xA;[libx264 @ 0x5e2ef8b01340] kb/s:8.27&#xA;Conversion failed!&#xA;Traceback (most recent call last):&#xA;  File "/home/avishka/projects/read-process-stream/minimal-ffmpeg-error.py", line 58, in <module>&#xA;    process.stdin.write(image_bytes)&#xA;BrokenPipeError: [Errno 32] Broken pipe&#xA;</module>

  


• ffmpeg piped output producing incorrect metadata frame count

  8 décembre 2024, par Xorgon

  The short version : Using piped output from ffmpeg produces a file with incorrect metadata.

  


  ffmpeg -y -i .\test_mp4.mp4 -f avi -c:v libx264 - > output.avi to make an AVI file using the pipe output.

  


  ffprobe -v error -count_frames -show_entries stream=duration,nb_read_frames,r_frame_rate .\output.avi

  


  The output will show that the metadata does not match the actual frames contained in the video.

  


  Details below.

  


  ---

  


  Using Python, I am attempting to use ffmpeg to compress videos and put them in a PowerPoint. This works great, however, the video files themselves have incorrect frame counts which can cause issues when I read from those videos in other code.

  


  Edit for clarification : by "frame count" I mean the metadata frame count. The actual number of frames contained in the video is correct, but querying the metadata gives an incorrect frame count.

  


  Having eliminated the PowerPoint aspect of the code, I've narrowed this down to the following minimal reproducing example of saving an output from an ffmpeg pipe :

  


  from subprocess import Popen, PIPE

video_path = 'test_mp4.mp4'

ffmpeg_pipe = Popen(['ffmpeg',
                     '-y',  # Overwrite files
                     '-i', f'{video_path}',  # Input from file
                     '-f', 'avi',  # Output format
                     '-c:v', 'libx264',  # Codec
                     '-'],  # Output to pipe
                    stdout=PIPE)

new_path = "piped_video.avi"
vid_file = open(new_path, "wb")
vid_file.write(ffmpeg_pipe.stdout.read())
vid_file.close()


  


  I've tested several different videos. One small example video that I've tested can be found here.

  


  I've tried a few different codecs with avi format and tried libvpx with webm format. For the avi outputs, the frame count usually reads as 1073741824 (2^30). Weirdly, for the webm format, the frame count read as -276701161105643264.

  


  Edit : This issue can also be reproduced with just ffmpeg in command prompt using the following command :
ffmpeg -y -i .\test_mp4.mp4 -f avi -c:v libx264 - > output.avi

  


  This is a snippet I used to read the frame count, but one could also see the error by opening the video details in Windows Explorer and seeing the total time as something like 9942 hours, 3 minutes, and 14 seconds.

  


  import cv2

video_path = 'test_mp4.mp4'
new_path = "piped_video.webm"

cap = cv2.VideoCapture(video_path)
print(f"Original video frame count: = {int(cap.get(cv2.CAP_PROP_FRAME_COUNT)):d}")
cap.release()

cap = cv2.VideoCapture(new_path)
print(f"Piped video frame count: = {int(cap.get(cv2.CAP_PROP_FRAME_COUNT)):d}")
cap.release()


  


  The error can also be observed using ffprobe with the following command : ffprobe -v error -count_frames -show_entries stream=duration,nb_read_frames,r_frame_rate .\output.avi. Note that the frame rate and number of frames counted by ffprobe do not match with the duration from the metadata.

  


  For completeness, here is the ffmpeg output :

  


  ffmpeg version 2023-06-11-git-09621fd7d9-full_build-www.gyan.dev Copyright (c) 2000-2023 the FFmpeg developers
  built with gcc 12.2.0 (Rev10, Built by MSYS2 project)
  configuration: --enable-gpl --enable-version3 --enable-static --disable-w32threads --disable-autodetect --enable-fontconfig --enable-iconv --enable-gnutls --enable-libxml2 --enable-gmp --enable-bzlib --enable-lzma --enable-libsnappy --enable-zlib --enable-librist --enable-libsrt --enable-libssh --enable-libzmq --enable-avisynth --enable-libbluray --enable-libcaca --enable-sdl2 --enable-libaribb24 --enable-libaribcaption --enable-libdav1d --enable-libdavs2 --enable-libuavs3d --enable-libzvbi --enable-librav1e --enable-libsvtav1 --enable-libwebp --enable-libx264 --enable-libx265 --enable-libxavs2 --enable-libxvid --enable-libaom --enable-libjxl --enable-libopenjpeg --enable-libvpx --enable-mediafoundation --enable-libass --enable-frei0r --enable-libfreetype --enable-libfribidi --enable-liblensfun --enable-libvidstab --enable-libvmaf --enable-libzimg --enable-amf --enable-cuda-llvm --enable-cuvid --enable-ffnvcodec --enable-nvdec --enable-nvenc --enable-d3d11va --enable-dxva2 --enable-libvpl --enable-libshaderc --enable-vulkan --enable-libplacebo --enable-opencl --enable-libcdio --enable-libgme --enable-libmodplug --enable-libopenmpt --enable-libopencore-amrwb --enable-libmp3lame --enable-libshine --enable-libtheora --enable-libtwolame --enable-libvo-amrwbenc --enable-libcodec2 --enable-libilbc --enable-libgsm --enable-libopencore-amrnb --enable-libopus --enable-libspeex --enable-libvorbis --enable-ladspa --enable-libbs2b --enable-libflite --enable-libmysofa --enable-librubberband --enable-libsoxr --enable-chromaprint
  libavutil      58. 13.100 / 58. 13.100
  libavcodec     60. 17.100 / 60. 17.100
  libavformat    60.  6.100 / 60.  6.100
  libavdevice    60.  2.100 / 60.  2.100
  libavfilter     9.  8.101 /  9.  8.101
  libswscale      7.  3.100 /  7.  3.100
  libswresample   4. 11.100 /  4. 11.100
  libpostproc    57.  2.100 / 57.  2.100
Input #0, mov,mp4,m4a,3gp,3g2,mj2, from 'test_mp4.mp4':
  Metadata:
    major_brand     : mp42
    minor_version   : 0
    compatible_brands: isommp42
    creation_time   : 2022-08-10T12:54:09.000000Z
  Duration: 00:00:06.67, start: 0.000000, bitrate: 567 kb/s
  Stream #0:0[0x1](eng): Video: h264 (High) (avc1 / 0x31637661), yuv420p(progressive), 384x264 [SAR 1:1 DAR 16:11], 563 kb/s, 30 fps, 30 tbr, 30k tbn (default)
    Metadata:
      creation_time   : 2022-08-10T12:54:09.000000Z
      handler_name    : Mainconcept MP4 Video Media Handler
      vendor_id       : [0][0][0][0]
      encoder         : AVC Coding
Stream mapping:
  Stream #0:0 -> #0:0 (h264 (native) -> h264 (libx264))
Press [q] to stop, [?] for help
[libx264 @ 0000018c68c8b9c0] using SAR=1/1
[libx264 @ 0000018c68c8b9c0] using cpu capabilities: MMX2 SSE2Fast SSSE3 SSE4.2 AVX FMA3 BMI2 AVX2
[libx264 @ 0000018c68c8b9c0] profile High, level 2.1, 4:2:0, 8-bit
Output #0, avi, to 'pipe:':
  Metadata:
    major_brand     : mp42
    minor_version   : 0
    compatible_brands: isommp42
    ISFT            : Lavf60.6.100
  Stream #0:0(eng): Video: h264 (H264 / 0x34363248), yuv420p(progressive), 384x264 [SAR 1:1 DAR 16:11], q=2-31, 30 fps, 30 tbn (default)
    Metadata:
      creation_time   : 2022-08-10T12:54:09.000000Z
      handler_name    : Mainconcept MP4 Video Media Handler
      vendor_id       : [0][0][0][0]
      encoder         : Lavc60.17.100 libx264
    Side data:
      cpb: bitrate max/min/avg: 0/0/0 buffer size: 0 vbv_delay: N/A
[out#0/avi @ 0000018c687f47c0] video:82kB audio:0kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: 3.631060%
frame=  200 fps=0.0 q=-1.0 Lsize=      85kB time=00:00:06.56 bitrate= 106.5kbits/s speed=76.2x    
[libx264 @ 0000018c68c8b9c0] frame I:1     Avg QP:16.12  size:  3659
[libx264 @ 0000018c68c8b9c0] frame P:80    Avg QP:21.31  size:   647
[libx264 @ 0000018c68c8b9c0] frame B:119   Avg QP:26.74  size:   243
[libx264 @ 0000018c68c8b9c0] consecutive B-frames:  3.0% 53.0%  0.0% 44.0%
[libx264 @ 0000018c68c8b9c0] mb I  I16..4: 17.6% 70.6% 11.8%
[libx264 @ 0000018c68c8b9c0] mb P  I16..4:  0.8%  1.7%  0.6%  P16..4: 17.6%  4.6%  3.3%  0.0%  0.0%    skip:71.4%
[libx264 @ 0000018c68c8b9c0] mb B  I16..4:  0.1%  0.3%  0.2%  B16..8: 11.7%  1.4%  0.4%  direct: 0.6%  skip:85.4%  L0:32.0% L1:59.7% BI: 8.3%
[libx264 @ 0000018c68c8b9c0] 8x8 transform intra:59.6% inter:62.4%
[libx264 @ 0000018c68c8b9c0] coded y,uvDC,uvAC intra: 48.5% 0.0% 0.0% inter: 3.5% 0.0% 0.0%
[libx264 @ 0000018c68c8b9c0] i16 v,h,dc,p: 19% 39% 25% 17%
[libx264 @ 0000018c68c8b9c0] i8 v,h,dc,ddl,ddr,vr,hd,vl,hu: 21% 25% 30%  3%  3%  4%  4%  4%  5%
[libx264 @ 0000018c68c8b9c0] i4 v,h,dc,ddl,ddr,vr,hd,vl,hu: 22% 20% 16%  6%  8%  8%  8%  5%  6%
[libx264 @ 0000018c68c8b9c0] i8c dc,h,v,p: 100%  0%  0%  0%
[libx264 @ 0000018c68c8b9c0] Weighted P-Frames: Y:0.0% UV:0.0%
[libx264 @ 0000018c68c8b9c0] ref P L0: 76.2%  7.9% 11.2%  4.7%
[libx264 @ 0000018c68c8b9c0] ref B L0: 85.6% 12.9%  1.5%
[libx264 @ 0000018c68c8b9c0] ref B L1: 97.7%  2.3%
[libx264 @ 0000018c68c8b9c0] kb/s:101.19


  


  So the question is : why does this happen, and how can one avoid it ?

  


• pyqt5 gui dependent on ffmpeg compiled with pyinstaller doesn't run on other machines ?

  19 octobre 2022, par Soren

  I am trying to create a simple Pyqt5 GUI for Windows 10 that uses OpenAI's model Whisper to transcribe a sound file and outputting the results in an Excel-file. It works on my own computer where I have installed the necessary dependencies for Whisper as stated on their github i.e. FFMEG. I provide a minimal example of my code below :

  


  # Import library
import whisper
import os
from PyQt5 import QtCore, QtGui, QtWidgets
import pandas as pd
import xlsxwriter


class Ui_Dialog(QtWidgets.QDialog):
    
    
    # Define functions to use in GUI
   
    # Define function for selecting input files
    def browsefiles(self, Dialog):
      
       
       # Make Dialog box and save files into tuple of paths
       files = QtWidgets.QFileDialog().getOpenFileNames(self, "Select soundfiles", os.getcwd(), "lyd(*mp2 *.mp3 *.mp4 *.m4a *wma *wav)")
       
       self.liste = []
       for url in range(len(files[0])):
           self.liste.append(files[0][url])   

    
    def model_load(self, Dialog):
               
        # Load picked model
        self.model = whisper.load_model(r'C:\Users\Søren\Downloads\Whisper_gui\models' + "\\" + self.combo_modelSize.currentText() + ".pt") ##the path is set to where the models are on the other machine
        
    
    def run(self, Dialog):
                
        # Make list for sound files
        liste_df = []
        
        
        # Running loop for interpreting and encoding sound files
        for url in range(len(self.liste)):
                          
            # Make dataframe
            df = pd.DataFrame(columns=["filename", "start", "end", "text"])
            
            # Run model
            result = self.model.transcribe(self.liste[url])
                            
            # Extract results
            for i in range(len(result["segments"])):
                start = result["segments"][i]["start"]
                end = result["segments"][i]["end"]
                text = result["segments"][i]["text"]
                
                df = df.append({"filename": self.liste[url].split("/")[-1],
                            "start": start, 
                            "end": end, 
                            "text": text}, ignore_index=True)
            
            # Add detected language to dataframe
            df["sprog"] = result["language"]
            
            
            liste_df.append(df)
        
        
        
        # Make excel output
        
        # Concatenate list of dfs
        dataframe = pd.concat(liste_df)
        
        
        # Create a Pandas Excel writer using XlsxWriter as the engine.
        writer = pd.ExcelWriter(self.liste[0].split(".")[0] + '_OUTPUT.xlsx', engine='xlsxwriter')
        writer_wrap_format = writer.book.add_format({"text_wrap": True, 'num_format': '@'})


        # Write the dataframe data to XlsxWriter. Turn off the default header and
        # index and skip one row to allow us to insert a user defined header.
        dataframe.to_excel(writer, sheet_name="Output", startrow=1, header=False, index=False)

        # Get the xlsxwriter workbook and worksheet objects.
        #workbook = writer.book
        worksheet = writer.sheets["Output"]

        # Get the dimensions of the dataframe.
        (max_row, max_col) = dataframe.shape

        # Create a list of column headers, to use in add_table().
        column_settings = [{'header': column} for column in dataframe.columns]

        # Add the Excel table structure. Pandas will add the data.
        worksheet.add_table(0, 0, max_row, max_col - 1, {'columns': column_settings})

        # Make the columns wider for clarity.
        worksheet.set_column(0, max_col - 1, 12)
        
        in_col_no = xlsxwriter.utility.xl_col_to_name(dataframe.columns.get_loc("text"))
        
        worksheet.set_column(in_col_no + ":" + in_col_no, 30, writer_wrap_format)

        # Close the Pandas Excel writer and output the Excel file.
        writer.save()
        writer.close()
    
    
    ## Design setup
    
    def setupUi(self, Dialog):
        Dialog.setObjectName("Dialog")
        Dialog.resize(730, 400)
        
        self.select_files = QtWidgets.QPushButton(Dialog)
        self.select_files.setGeometry(QtCore.QRect(40, 62, 81, 31))
        font = QtGui.QFont()
        font.setPointSize(6)
        self.select_files.setFont(font)
        self.select_files.setObjectName("select_files")
        
    
               
        
        self.combo_modelSize = QtWidgets.QComboBox(Dialog)
        self.combo_modelSize.setGeometry(QtCore.QRect(40, 131, 100, 21))
        font = QtGui.QFont()
        font.setPointSize(6)
        self.combo_modelSize.setFont(font)
        self.combo_modelSize.setObjectName("combo_modelSize")
               
        
        self.runButton = QtWidgets.QPushButton(Dialog)
        self.runButton.setGeometry(QtCore.QRect(40, 289, 71, 21))
        font = QtGui.QFont()
        font.setPointSize(6)
        self.runButton.setFont(font)
        self.runButton.setObjectName("runButton")
        
        
       

        self.retranslateUi(Dialog)
        QtCore.QMetaObject.connectSlotsByName(Dialog)
        
        
        
        modelSize_options = ['Chose model', 'tiny', 'base', 'small', 'medium', 'large']
        self.combo_modelSize.addItems(modelSize_options)
        
        # Do an action!
        self.select_files.clicked.connect(self.browsefiles)
        self.combo_modelSize.currentIndexChanged.connect(self.model_load)
        self.runButton.clicked.connect(self.run)
        
        
        
    

    def retranslateUi(self, Dialog):
        _translate = QtCore.QCoreApplication.translate
        Dialog.setWindowTitle(_translate("Dialog", "Dialog"))
        self.runButton.setText(_translate("Dialog", "Go!"))
        self.select_files.setText(_translate("Dialog", "Select"))


if __name__ == "__main__":
    import sys
    app = QtWidgets.QApplication(sys.argv)
    Dialog = QtWidgets.QDialog()
    ui = Ui_Dialog()
    ui.setupUi(Dialog)
    Dialog.show()
    sys.exit(app.exec_())


  


  I compile this app with pyinstaller using the following code. I had some issues to begin with so I found other with similar problems and ended up with this :

  


  pyinstaller --onedir --hidden-import=pytorch --collect-data torch --copy-metadata torch --copy-metadata tqdm --copy-metadata tokenizers --copy-metadata importlib_metadata --hidden-import="sklearn.utils._cython_blas" --hidden-import="sklearn.neighbors.typedefs" --hidden-import="sklearn.neighbors.quad_tree" --hidden-import="sklearn.tree" --hidden-import="sklearn.tree._utils" --copy-metadata regex --copy-metadata requests --copy-metadata packaging --copy-metadata filelock --copy-metadata numpy --add-data "./ffmpeg/*;./ffmpeg/" --hidden-import=whisper --copy-metadata whisper --collect-data whisper minimal_example_whisper.py

  


  When I take the outputtet dist directory and try to run the app on another Windows machine without FFMPEG installed (or Whisper or any other things), I get the following error from the terminal as I push the "run" button in the app (otherwise the app does run).

  


  C:\Users\Søren>"G:\minimal_example_whisper\minimal_example_whisper.exe"
whisper\transcribe.py:70: UserWarning: FP16 is not supported on CPU; using FP32 instead
Traceback (most recent call last):
  File "minimal_example_whisper.py", line 45, in run
  File "whisper\transcribe.py", line 76, in transcribe
  File "whisper\audio.py", line 111, in log_mel_spectrogram
  File "whisper\audio.py", line 42, in load_audio
  File "ffmpeg\_run.py", line 313, in run
  File "ffmpeg\_run.py", line 284, in run_async
  File "subprocess.py", line 951, in __init__
  File "subprocess.py", line 1420, in _execute_child
FileNotFoundError: [WinError 2] Den angivne fil blev ikke fundet


  


  I suspect this has something to do with FFMPEG not being installed on the other machines system ? Does anyone have an automatic solution for this when compiling the app or can it simply only run on machines that has FFMPEG installed ?

  


  Thanks in advance !