Update app.py

app.py

@@ -13,14 +13,6 @@ decoder = BeamSearchDecoderCTC.load_from_hf_hub(lmID)
 p = pipeline("automatic-speech-recognition", model="aware-ai/robust-wav2vec2-base-german-lowercase", decoder=decoder)
 ttp = pipeline("text2text-generation", model="aware-ai/marian-german-grammar")
 
-vadmodel, utils = torch.hub.load(repo_or_dir='snakers4/silero-vad',
-                              model='silero_vad',
-                              force_reload=False)
-
-(get_speech_timestamps,
- _, read_audio,
- *_) = utils
-
 #model = M2M100ForConditionalGeneration.from_pretrained("facebook/m2m100_1.2B")
 #tokenizer = M2M100Tokenizer.from_pretrained("facebook/m2m100_1.2B")
 
@@ -36,28 +28,11 @@ def translate(src, tgt, text):
     return result
 
 def transcribe(audio):
-    log = ""
-    sampling_rate = 16000
-    start_time = time.time()
-    audio, sr = librosa.load(audio, sr=sampling_rate)
-    log += "--- %s seconds audio loading ---" + str(time.time() - start_time)
-    start_time = time.time()
-    speech_timestamps = get_speech_timestamps(audio, vadmodel, sampling_rate=sampling_rate)
-    log += "\n--- %s seconds audio timestamps---" + str(time.time() - start_time)
-    start_time = time.time()
-    chunks = [audio[i["start"]:i["end"]] for i in speech_timestamps]
-    log += "\n--- %s seconds audio chunking---" + str(time.time() - start_time)
-    start_time = time.time()
-    transcribed = " ".join([text["text"] for text in p(chunks, chunk_length_s=20, stride_length_s=(0, 0))])
-    log += "\n--- %s seconds audio transcription ---" + str(time.time() - start_time)
-    start_time = time.time()    
+    transcribed = p(audio, chunk_length_s=20, stride_length_s=(0, 0))["text"]
+    
     punctuated = ttp(transcribed, max_length = 512)[0]["generated_text"]
-    log += "\n--- %s seconds audio formatting ---" + str(time.time() - start_time)
-    start_time = time.time()
-    p(audio, chunk_length_s=20, stride_length_s=(0, 0))
-    log += "\n--- %s seconds full asr ---" + str(time.time() - start_time)
     
-    return transcribed, punctuated, log
+    return transcribed, punctuated
 
 def get_asr_interface():
     return gr.Interface(
@@ -66,7 +41,6 @@ def get_asr_interface():
             gr.inputs.Audio(source="microphone", type="filepath")
         ],
         outputs=[
-            "textbox",
             "textbox",
             "textbox"
         ])