Support using Vosk instead of Whisper for transcripts

For Esperanto, Vosk is used automatically, since Whisper doesn't
support Esperanto

lib/args.py

@@ -11,6 +11,7 @@ class Options:
     filter = None
     episodes = 'new'
     numEpisodes = 1
+    tts = 'whisper'
 
 def outputHelp():
     print("Usage: ./process [args] [filter [age] [episodes]]")
@@ -26,6 +27,8 @@ def outputHelp():
     print("    Skip generating transcripts and translations of podcast episodes")
     print("--verbose")
     print("    Output a detailed report of actions as they are being taken")
+    print("--vosk")
+    print("    Use Vosk instead of Whisper to generate transcripts")
     print("")
     print("filter:")
     print("    If specified, only configured podcasts which match the filter will be processed")
@@ -61,6 +64,8 @@ def read():
             opts.generate = False
         elif arg == "--verbose":
             opts.verbose = True
+        elif arg == '--vosk':
+            opts.tts = 'vosk'
         elif arg == 'new':
             opts.episodes = 'new'
         elif arg == 'old':

lib/files.py

@@ -109,7 +109,7 @@ def hasTranslation(audioFile):
     translationFile = getTranslationFilename(audioFile)
     return os.path.isfile(translationFile)
 
-def generateFromAudio(audioFile, task):
+def generateFromAudio(audioFile, task, options):
     if task == 'transcribe':
         newExt = 'transcript.txt'
     else:
@@ -119,14 +119,18 @@ def generateFromAudio(audioFile, task):
     langCode = getLangCode(audioFile)
     fileParts = audioFile.split('/')
     fileName = fileParts.pop()
+    newFileName = replaceExtension(fileName, newExt)
     dir = '/'.join(fileParts)
     os.chdir(dir)
 
-    cmd = f"whisper {fileName} --model medium --language {langCode} --task {task} --output_format vtt --fp16 False"
-    os.system(cmd)
+    if options.tts == 'vosk' or langCode == 'eo':
+        cmd = f"vosk-transcriber -i {fileName} -l {langCode} -t srt -o {newFileName}"
+        os.system(cmd)
+    else:
+        cmd = f"whisper {fileName} --model medium --language {langCode} --task {task} --output_format srt --fp16 False"
+        os.system(cmd)
 
-    # rename transcript/translation file generated
-    generatedFile = replaceExtension(fileName, 'vtt')
-    newFileName = replaceExtension(fileName, newExt)
-    os.rename(generatedFile, newFileName)
+        # rename transcript/translation file generated
+        generatedFile = replaceExtension(fileName, 'vtt')
+        os.rename(generatedFile, newFileName)
     os.chdir('../../..')

process

@@ -30,11 +30,13 @@ for audioFile in audioFiles:
     if not files.hasTranscript(audioFile):
         skipped = False
         print(f"Transcribing audio: {audioFile} ({audioLength})", flush=True)
-        files.generateFromAudio(audioFile, 'transcribe')
-    if language in config['translate'] and not files.hasTranslation(audioFile):
+        files.generateFromAudio(audioFile, 'transcribe', options)
+    if (language in config['translate'] and not files.hasTranslation(audioFile)
+        and language != 'eo' and options.tts != 'vosk'
+    ):
         skipped = False
         print(f"Translating audio: {audioFile} ({audioLength})", flush=True)
-        files.generateFromAudio(audioFile, 'translate')
+        files.generateFromAudio(audioFile, 'translate', options)
     if skipped and options.verbose:
         print(f"Skipped audio: {audioFile} ({audioLength})")