Update app.py

app.py

@@ -2,43 +2,35 @@ from transformers import pipeline
 import gradio as gr
 from pyctcdecode import BeamSearchDecoderCTC
 
-
-lmID = "aware-ai/german-lowercase-4gram-kenlm"
+lmID = "aware-ai/german-lowercase-5gram-kenlm"
 decoder = BeamSearchDecoderCTC.load_from_hf_hub(lmID)
-p = pipeline("automatic-speech-recognition", model="aware-ai/robust-wav2vec2-base-german-lowercase", decoder=decoder)
+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")
 
 def transcribe(audio):
-    transcribed = p(audio[1], chunk_length_s=20, stride_length_s=(0, 0))["text"]  
-    return transcribed
+    transcribed = p(audio, chunk_length_s=16, stride_length_s=(4, 0))["text"]
+    
+    punctuated = ttp(transcribed, max_length = 512)[0]["generated_text"]
     
-def punctuate(text):
-    punctuated = ttp(text, max_length = 512)[0]["generated_text"]
-    return punctuated 
+    return transcribed, punctuated 
 
 def get_asr_interface():
     return gr.Interface(
         fn=transcribe, 
         inputs=[
-            gr.inputs.Audio(source="microphone")
+            gr.inputs.Audio(source="microphone", type="filepath")
         ],
         outputs=[
             "textbox",
-        ])
-        
-def get_punctuate_interface():
-    return gr.Interface(
-        fn=punctuate, 
-        inputs=[
             "textbox"
-        ],
-        outputs=[
-            "textbox",
         ])
 
 interfaces = [
     get_asr_interface(),
-    get_punctuate_interface(),
 ]
 
-gr.Series(get_asr_interface(),get_punctuate_interface()).launch(server_name = "0.0.0.0")
+names = [
+    "ASR",
+]
+
+gr.TabbedInterface(interfaces, names).launch(server_name = "0.0.0.0")