README.md

@@ -14,8 +14,8 @@ add "transcribe" to LOCAL_APPS in local_setttings.py
     apt install portaudio19-dev
     python3 -m venv venv /opt/whisper-timestamped/
     /opt/whisper-timestamped/bin/pip install \
-     torch==2.3.1+cpu \
-     torchaudio==2.3.1+cpu \
+     torch==1.13.1+cpu \
+     torchaudio==0.13.1+cpu \
      -f https://download.pytorch.org/whl/torch_stable.html
     /opt/whisper-timestamped/bin/pip install whisper-timestamped transformers auditok
 

run_whisper.py

@@ -6,9 +6,8 @@ import sys
 import whisper_timestamped as whisper
 
 
-def transcribe(path, language=None, translate=None, model="small"):
-    #model = whisper.load_model("large-v3-turbo")
-    model = whisper.load_model(model)
+def transcribe(path, language=None, translate=None):
+    model = whisper.load_model("large-v3-turbo")
     audio = whisper.load_audio(path)
     if translate:
         task = "translate"
@@ -34,11 +33,10 @@ if __name__ == "__main__":
     usage = "usage: %(prog)s [options] path output"
     parser = ArgumentParser(usage=usage, prog="run_whisper")
     parser.add_argument("-l", "--language", dest="language", default=None)
-    parser.add_argument("-m", "--model", dest="model", default="small")
-    parser.add_argument("-t", "--translate", dest="translate", action='store_true', default=False)
+    parser.add_argument("-t", "--translate", dest="translate", default=None)
     parser.add_argument("path", metavar="path", type=str, help="media input")
     parser.add_argument("output", metavar="output", type=str, help="json output")
     opts = parser.parse_args()
-    result = transcribe(opts.path, opts.language, opts.translate, opts.model)
+    result = transcribe(opts.path, opts.language, opts.translate)
     with open(opts.output, "w") as fd:
         json.dump(result, fd, indent=2, ensure_ascii=False)

transcribe.py

@@ -1,4 +1,3 @@
-import json
 import logging
 import os
 import shutil
@@ -94,10 +93,12 @@ def prepare_joint_annotations(result, target_length=200):
 
 def extract_subtitles(item, user, layer, translate, gpu=False, join_sentences=False):
     language = None
+    """
     if "language" not in item.data:
         language = None
     else:
         language = ox.iso.langTo2Code(item.data["language"][0])
+    """
     if not item.streams():
         logger.error("skip item without media %s: %s", item.public_id)
         return False