Update app.py

app.py

@@ -7,6 +7,8 @@ from pyctcdecode import BeamSearchDecoderCTC
 lmID = "aware-ai/german-lowercase-5gram-kenlm"
 decoder = BeamSearchDecoderCTC.load_from_hf_hub(lmID)
 p = pipeline("automatic-speech-recognition", model="aware-ai/robust-wav2vec2-xls-r-300m-german-lowercase", decoder=decoder)
+ttp = pipeline("text2text-generation", model="aware-ai/marian-german-grammar")
+
 
 model = M2M100ForConditionalGeneration.from_pretrained("facebook/m2m100_1.2B")
 tokenizer = M2M100Tokenizer.from_pretrained("facebook/m2m100_1.2B")
@@ -25,7 +27,9 @@ def translate(src, tgt, text):
 def transcribe(audio):
     transcribed = p(audio, chunk_length_s=10, stride_length_s=(4, 2))["text"]
     
-    return transcribed
+    punctuated = ttp(transcribed, max_length = 512)[0]["generated_text"]
+    
+    return transcribed, punctuated 
 
 def get_asr_interface():
     return gr.Interface(
@@ -34,6 +38,7 @@ def get_asr_interface():
             gr.inputs.Audio(source="microphone", type="filepath")
         ],
         outputs=[
+            "textbox",
             "textbox"
         ])