|
| 1 | +from gooey import Gooey, GooeyParser |
| 2 | +import subprocess |
| 3 | +import os |
| 4 | +import re |
| 5 | + |
| 6 | +@Gooey(program_name='whisper.cppGUI', |
| 7 | + menu=[{'name': 'File', |
| 8 | + 'items': [{ |
| 9 | + 'type': 'AboutDialog', |
| 10 | + 'menuTitle': 'Download GGML models', |
| 11 | + 'name': 'Download the GGML models below', |
| 12 | + 'website': 'https://huggingface.co/datasets/ggerganov/whisper.cpp/tree/main' |
| 13 | + },{ |
| 14 | + 'type': 'Link', |
| 15 | + 'menuTitle': 'Visit Our Site', |
| 16 | + 'url': 'https://github.com/Topping1/whispercppGUI' |
| 17 | + }] |
| 18 | + }] |
| 19 | + ) |
| 20 | +def main(): |
| 21 | + parser = GooeyParser(description='GUI for whisper.cpp, a high-performance C++ port of OpenAI\'s Whisper') |
| 22 | + |
| 23 | + parser.add_argument( |
| 24 | + '--file', |
| 25 | + metavar='File to transcribe', |
| 26 | + help='supports all file types supported by FFMPEG', |
| 27 | + widget='FileChooser') |
| 28 | + |
| 29 | + parser.add_argument( |
| 30 | + '--model', |
| 31 | + metavar='GGML model (.bin)', |
| 32 | + help='select GGML model (tiny,base,small,medium,large)', |
| 33 | + widget='FileChooser') |
| 34 | + |
| 35 | + parser.add_argument( |
| 36 | + '--language', |
| 37 | + metavar='Language', |
| 38 | + help='select language for transcription', |
| 39 | + choices=['en','zh','de','es','ru','ko','fr','ja','pt','tr','pl','ca','nl','ar','sv','it','id','hi','fi','vi','iw','uk','el','ms','cs','ro','da','hu','ta','no','th','ur','hr','bg','lt','la','mi','ml','cy','sk','te','fa','lv','bn','sr','az','sl','kn','et','mk','br','eu','is','hy','ne','mn','bs','kk','sq','sw','gl','mr','pa','si','km','sn','yo','so','af','oc','ka','be','tg','sd','gu','am','yi','lo','uz','fo','ht','ps','tk','nn','mt','sa','lb','my','bo','tl','mg','as','tt','haw','ln','ha','ba','jw','su'], |
| 40 | + widget='FilterableDropdown') |
| 41 | + |
| 42 | + parser.add_argument( |
| 43 | + '--translate', |
| 44 | + action='store_true', |
| 45 | + help='check to translate transcription to english') |
| 46 | + |
| 47 | + parser.add_argument( |
| 48 | + '-otxt', |
| 49 | + '--output-txt', |
| 50 | + action='store_true', |
| 51 | + help='check to output result in a text file') |
| 52 | + |
| 53 | + parser.add_argument( |
| 54 | + '-osrt', |
| 55 | + '--output-srt', |
| 56 | + action='store_true', |
| 57 | + help='check to output result in a srt file') |
| 58 | + |
| 59 | + parser.add_argument( |
| 60 | + '-ovtt', |
| 61 | + '--output-vtt', |
| 62 | + action='store_true', |
| 63 | + help='check to output result in a vtt file') |
| 64 | + |
| 65 | + parser.add_argument( |
| 66 | + '--speed-up2', |
| 67 | + action='store', |
| 68 | + default=1, |
| 69 | + help='alternative speed up based on FFMPEG. Type here the speed up factor (e.g. 1.5). This enables automatically SRT output with corrected timestamps', |
| 70 | + widget='DecimalField') |
| 71 | + |
| 72 | + parser.add_argument( |
| 73 | + '--others', |
| 74 | + action='store', |
| 75 | + default="", |
| 76 | + help='This textbox lets the user add other command line parameters that are not included in this GUI') |
| 77 | + |
| 78 | + parser.add_argument( |
| 79 | + '--shell', |
| 80 | + action='store_true', |
| 81 | + help='check to show the shell window instead of using the Gooey window. This can fix some UTF-8 errors') |
| 82 | + |
| 83 | + |
| 84 | + args = parser.parse_args() |
| 85 | +#enable for debugging |
| 86 | +# print(args) |
| 87 | +#pass args for later use in args=main() |
| 88 | + return args |
| 89 | + |
| 90 | + |
| 91 | + |
| 92 | +if __name__ == '__main__': |
| 93 | + |
| 94 | + |
| 95 | +# this section is inspired by |
| 96 | +# https://stackoverflow.com/questions/48767005/using-python-gooey-how-to-open-another-gui-after-clicking-one-out-of-multiple-bu |
| 97 | + |
| 98 | +#get arguments from main() for use here |
| 99 | + args=main() |
| 100 | + |
| 101 | +#workaround to process the arguments that evaluate to "True" or "False" |
| 102 | + |
| 103 | + if args.translate == True: |
| 104 | + arg_translate = "--translate" |
| 105 | + else: |
| 106 | + arg_translate = "" |
| 107 | + |
| 108 | + if args.output_txt == True: |
| 109 | + arg_out_txt = "--output-txt" |
| 110 | + else: |
| 111 | + arg_out_txt = "" |
| 112 | + |
| 113 | + if args.output_srt == True: |
| 114 | + arg_out_srt = "--output-srt" |
| 115 | + else: |
| 116 | + arg_out_srt = "" |
| 117 | + |
| 118 | + if args.output_vtt == True: |
| 119 | + arg_out_vtt = "--output-vtt" |
| 120 | + else: |
| 121 | + arg_out_vtt = "" |
| 122 | + |
| 123 | + |
| 124 | +#check if ffmpeg speedup was selected. If true, disable txt and vtt output |
| 125 | +#and disable whispercpp internal speed up. |
| 126 | + if float(args.speed_up2) != 1.0: |
| 127 | + arg_out_txt = "" |
| 128 | + arg_out_vtt = "" |
| 129 | + arg_speed = "" |
| 130 | + arg_out_srt = "--output-srt" |
| 131 | + |
| 132 | +# Split the input filename and extension |
| 133 | + input_basename, _ = os.path.splitext(os.path.basename(args.file)) |
| 134 | + |
| 135 | +# Construct new output filenames based on input filename |
| 136 | + output_wav = f"{input_basename}.wav" |
| 137 | + output_srt = f"{input_basename}.wav.srt" |
| 138 | + output_fixed_srt = f"{input_basename}_fixed.srt" |
| 139 | + |
| 140 | +#first we process the input file with ffmpeg |
| 141 | +#here we construct the command line for ffmpeg and apply the FFMPEG speedup IF selected |
| 142 | + if float(args.speed_up2) != 1.0: |
| 143 | + cmd = f"ffmpeg.exe -y -i \"{args.file}\" -ar 16000 -ac 1 -c:a pcm_s16le -af atempo={args.speed_up2} {output_wav}" |
| 144 | + else: |
| 145 | + cmd = f"ffmpeg.exe -y -i \"{args.file}\" -ar 16000 -ac 1 -c:a pcm_s16le {output_wav}" |
| 146 | + |
| 147 | + if args.shell == True: |
| 148 | + #here we call the program with extra parameters to capture ffmpeg output |
| 149 | + process=subprocess.Popen(cmd, text=True) |
| 150 | + else: |
| 151 | + #workaround required to show ffmpeg output in the Gooey window |
| 152 | + #reference https://github.com/chriskiehl/Gooey/issues/355 |
| 153 | + startupinfo = subprocess.STARTUPINFO() |
| 154 | + startupinfo.dwFlags |= subprocess.STARTF_USESHOWWINDOW |
| 155 | + #here we call the program with extra parameters to capture ffmpeg output |
| 156 | + process=subprocess.Popen(cmd, |
| 157 | + startupinfo=startupinfo, |
| 158 | + stdout=subprocess.PIPE, |
| 159 | + stdin=subprocess.PIPE, |
| 160 | + stderr=subprocess.STDOUT) |
| 161 | + #here we print ffmpeg output to the Gooey window |
| 162 | + for line in process.stdout: |
| 163 | + line1=line.decode('utf-8') |
| 164 | + print(line1.rstrip()) |
| 165 | + |
| 166 | + |
| 167 | + |
| 168 | + |
| 169 | +#here we run whisperCPP |
| 170 | +#here we construct the command line for whisperCPP |
| 171 | + cmd = f"main.exe -f {output_wav} -m {args.model} -l {args.language} {arg_translate} {arg_out_txt} {arg_out_srt} {arg_out_vtt} {args.others}" |
| 172 | + |
| 173 | + |
| 174 | + if args.shell == True: |
| 175 | + #here we call the program with extra parameters to capture ffmpeg output |
| 176 | + process=subprocess.Popen(cmd, text=True) |
| 177 | + else: |
| 178 | + |
| 179 | + #workaround required to show whisperCPP output in the Gooey window |
| 180 | + #reference https://github.com/chriskiehl/Gooey/issues/355 |
| 181 | + |
| 182 | + startupinfo = subprocess.STARTUPINFO() |
| 183 | + startupinfo.dwFlags |= subprocess.STARTF_USESHOWWINDOW |
| 184 | + #here we call the program with extra parameters to capture whisperCPP output |
| 185 | + process=subprocess.Popen(cmd, |
| 186 | + startupinfo=startupinfo, |
| 187 | + stdout=subprocess.PIPE, |
| 188 | + stdin=subprocess.PIPE, |
| 189 | + stderr=subprocess.STDOUT) |
| 190 | + #here we print whisperCPP output to the Gooey window |
| 191 | + for line in process.stdout: |
| 192 | + line1=line.decode('utf-8') |
| 193 | + print(line1.rstrip()) |
| 194 | + |
| 195 | +#this section fixes the timestamps of the SRT file if the FFMPEG speedup was selected |
| 196 | +# |
| 197 | + |
| 198 | + if float(args.speed_up2) != 1.0: |
| 199 | + speedup_factor = float(args.speed_up2) # assign the speedup factor |
| 200 | + with open(output_srt, "r") as file: # Open the input SRT file |
| 201 | + content = file.read() |
| 202 | + file.close() |
| 203 | + matches = re.findall(r"\d{2}:\d{2}:\d{2},\d{3}", content) # Use regular expressions to match timestamps in the SRT file |
| 204 | + for match in matches: # Multiply the timestamps by the speedup factor |
| 205 | + parts = match.split(":") |
| 206 | + hours = int(parts[0]) |
| 207 | + minutes = int(parts[1]) |
| 208 | + seconds = int(parts[2].split(",")[0]) |
| 209 | + milliseconds = int(parts[2].split(",")[1]) |
| 210 | + total_milliseconds = (hours * 3600 + minutes * 60 + seconds) * 1000 + milliseconds |
| 211 | + total_milliseconds *= speedup_factor |
| 212 | + new_hours = f'{int(total_milliseconds // 3600000):02d}' |
| 213 | + new_minutes = f'{int((total_milliseconds % 3600000) // 60000):02d}' |
| 214 | + new_seconds = f'{int(((total_milliseconds % 3600000) % 60000) // 1000):02d}' |
| 215 | + new_milliseconds = f'{int(((total_milliseconds % 3600000) % 60000) % 1000):03d}' |
| 216 | + new_time = f"{new_hours}:{new_minutes}:{new_seconds},{new_milliseconds}" |
| 217 | + content = content.replace(match, new_time) |
| 218 | + with open(output_fixed_srt, "w") as file: # Write the adjusted SRT file to the output file |
| 219 | + file.write(content) |
| 220 | + file.close() |
| 221 | + os.remove(output_srt) #remove output.srt temporary file |
| 222 | +#end of section that fixes the timestamps |
| 223 | + |
| 224 | +#remove output.wav temporary file created by ffmpeg |
| 225 | +os.remove(output_wav) |
0 commit comments