Update app.py

app.py

@@ -1,22 +1,26 @@
 import gradio as gr
-from transformers import pipeline, Wav2Vec2ForCTC, Wav2Vec2Processor
+from transformers import pipeline
+import numpy as np
 
-# Load the ASR model and processor
-model = Wav2Vec2ForCTC.from_pretrained("facebook/wav2vec2-base-960h")
-processor = Wav2Vec2Processor.from_pretrained("facebook/wav2vec2-base-960h")
+transcriber = pipeline("automatic-speech-recognition", model="facebook/xm_transformer_s2ut_hk-en")
 
-def transcribe(audio):
-    # Process the audio and transcribe using the Hugging Face ASR model
-    input_values = processor(audio, return_tensors="pt", sampling_rate=16000).input_values
-    logits = model(input_values).logits
-    predicted_ids = torch.argmax(logits, dim=-1)
-    transcription = processor.batch_decode(predicted_ids)
-    
-    return transcription[0]
+def transcribe(stream, new_chunk):
+    sr, y = new_chunk
+    y = y.astype(np.float32)
+    y /= np.max(np.abs(y))
 
-gr.Interface(
-    fn=transcribe,
-    inputs=gr.Audio(source="microphone", type="filepath"),
-    outputs="text",
-    live=True
-).launch()
+    if stream is not None:
+        stream = np.concatenate([stream, y])
+    else:
+        stream = y
+    return stream, transcriber({"sampling_rate": sr, "raw": stream})["text"]
+
+
+demo = gr.Interface(
+    transcribe,
+    ["state", gr.Audio(sources=["microphone"], streaming=True)],
+    ["state", "text"],
+    live=True,
+)
+
+demo.launch()