Update app.py

app.py

@@ -28,7 +28,7 @@ device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 print(f"Bruker enhet: {device}")
 
 @spaces.GPU(duration=60 * 2)
-def pipe(file, return_timestamps=False,lang):
+def pipe(file, return_timestamps=False,lang="no"):
     asr = pipeline(
         task="automatic-speech-recognition",
         model=MODEL_NAME,
@@ -71,10 +71,10 @@ def transcribe(file, return_timestamps=False,lang_nn=False):
         
     if not lang_nn:    
         if not return_timestamps:
-            text = pipe(file_to_transcribe, return_timestams=False, lang="no")["text"]
+            text = pipe(file_to_transcribe)["text"]
             formatted_text = format_output(text)
         else:
-            chunks = pipe(file_to_transcribe, return_timestamps=True, lang="no")["chunks"]
+            chunks = pipe(file_to_transcribe, return_timestamps=True)["chunks"]
             text = []
             for chunk in chunks:
                 start_time = time.strftime('%H:%M:%S', time.gmtime(chunk["timestamp"][0])) if chunk["timestamp"][0] is not None else "??:??:??"
@@ -84,7 +84,7 @@ def transcribe(file, return_timestamps=False,lang_nn=False):
             formatted_text = "<br>".join(text)
     else:
         if not return_timestamps:
-            text = pipe(file_to_transcribe, return_timestams=False,lang="nn")["text"]
+            text = pipe(file_to_transcribe,lang="nn")["text"]
             formatted_text = format_output(text)
         else:
             chunks = pipe(file_to_transcribe, return_timestamps=True,lang="nn")["chunks"]
@@ -95,7 +95,11 @@ def transcribe(file, return_timestamps=False,lang_nn=False):
                 line = f"[{start_time} -> {end_time}] {chunk['text']}"
                 text.append(line)
             formatted_text = "<br>".join(text)
-
+            
+    output_file = "transcription.txt"
+    with open(output_file, "w") as f:
+        f.write(re.sub('<br>', '\n', formatted_text))
+        
     if truncated:
         link="https://github.com/NbAiLab/nostram/blob/main/leverandorer.md"
         disclaimer = (
@@ -109,7 +113,7 @@ def transcribe(file, return_timestamps=False,lang_nn=False):
     formatted_text += "<br><br><i>Transkribert med NB-Whisper demo</i>"
 
     
-    return formatted_text
+    return formatted_text, output_file
 
 def _return_yt_html_embed(yt_url):
     video_id = yt_url.split("?v=")[-1]
@@ -162,7 +166,11 @@ with demo:
             gr.components.Checkbox(label="Inkluder tidsstempler"),
             gr.components.Checkbox(label="Nynorsk"),
         ],
-        outputs=gr.HTML(label="text"),
+        
+        outputs=[
+            gr.HTML(label="text"),
+            gr.File(label="Last ned transkripsjon")
+        ],
         #outputs="text",
         
         description=(