This seems to be good? hmmm

download.py

@@ -173,31 +173,39 @@ def caption_generator(social_media_url,uid, language="Autodetect", model_size=mo
       print(f"Starting whisper transcribe with {uid}.mp3")
       transcribe_whisper_result = transcribe(audio, translate_action=False, language=language, override_model_size=model_size)
       detected_language = LANGUAGES[transcribe_whisper_result["language"]]
-      translate_whisper_result = transcribe(audio, translate_action=True, language=detected_language, override_model_size=model_size)
       print(f"Transcribe successful!, writing files")
+
       vtt_path = tempdir / f"{transcribe_whisper_result['language']}.vtt"
-      en_vtt_path = tempdir / f"en.vtt"
 
       with open(vtt_path.resolve(), "w", encoding="utf-8") as vtt:
         write_vtt(transcribe_whisper_result["segments"], file=vtt)
 
-      with open(en_vtt_path.resolve(), "w", encoding="utf-8") as en_vtt:
-        write_vtt(transcribe_whisper_result["segments"], file=en_vtt)
+      whisper_result_captions = [
+        {
+          "language_tag": transcribe_whisper_result["language"],
+          "vtt_file": anvil.BlobMedia(content_type="text/plain", content=vtt_path.read_bytes(),
+                                      name=f"{uid}.{transcribe_whisper_result['language']}.vtt")
+        },
+      ]
+
+      if detected_language != "en":
+        print(f"Transcribe successful! Starting translation to English")
+        translate_whisper_result = transcribe(audio, translate_action=True, language=detected_language, override_model_size=model_size)
+        print(f"Trfan!, writing files")
+        en_vtt_path = tempdir / f"en.vtt"
+        with open(en_vtt_path.resolve(), "w", encoding="utf-8") as en_vtt:
+          write_vtt(transcribe_whisper_result["segments"], file=en_vtt)
+        print(f"Finished translation to English, preparing subtitle files")
+        whisper_result_captions.append(
+          {
+            "language_tag": "en",
+            "vtt_file": anvil.BlobMedia(content_type="text/plain", content=vtt_path.read_bytes(), name=f"{uid}.en.vtt")
+          }
+        )
 
     except Exception as e:
       print(f"Could not transcribe file: {e}")
-      return
-
-    whisper_result_captions =  [
-      {
-        "language_tag": transcribe_whisper_result["language"],
-        "vtt_file": anvil.BlobMedia(content_type="text/plain", content=vtt_path.read_bytes(), name=f"{uid}.{transcribe_whisper_result['language']}.vtt")
-      },
-      {
-        "language_tag": "en",
-        "vtt_file": anvil.BlobMedia(content_type="text/plain", content=vtt_path.read_bytes(), name=f"{uid}.en.vtt")
-      }
-    ]
+      raise
 
     return 'success', whisper_result_captions
 
@@ -315,7 +323,7 @@ def transcribe(audio, translate_action=True, language='Autodetect', override_mod
   }
 
   if language != 'Autodetect':
-    props["language"] = TO_LANGUAGE_CODE[language.lower()]
+    props["language"] = TO_LANGUAGE_CODE[language.lower()] if len(language) > 2 else language
 
   output = model.transcribe(audio, verbose=True, **props)
 

utils/apis.py

@@ -63,8 +63,8 @@ def test_api(url=''):
   # TODO: add an anvil server pingback to show we completed the queue operation
   return f"I've slept for 15 seconds and now I'm done. "
 
-#TODO: add telegram error handler here
-def caption(downloadable_url="",uid="", language="Autodetect", override_model_size=""):
+
+def caption(downloadable_url="", uid="", language="Autodetect", override_model_size=""):
   """
   :param media_id: The twitter media ID object
   :param user_id_str: The twitter user ID string