Create app.py

app.py

@@ -0,0 +1,30 @@
+# app.py
+
+import gradio as gr
+import torch
+import whisper
+
+# Load a fast Whisper model
+model = whisper.load_model("small")  # You can use "tiny" if you want even faster
+
+def transcribe_audio(audio):
+    # Audio is received as a tuple (sample_rate, numpy_array)
+    audio = audio[1]  # Get the raw audio waveform
+
+    # Whisper expects 16000 Hz sample rate
+    result = model.transcribe(audio, fp16=torch.cuda.is_available())
+    text = result["text"]
+    return text
+
+# Gradio Interface
+iface = gr.Interface(
+    fn=transcribe_audio,
+    inputs=gr.Audio(source="microphone", type="numpy", streaming=True),
+    outputs=gr.Textbox(label="Recognized Text"),
+    live=True,  # Important for real-time streaming
+    title="Real-time Voice to Text",
+    description="Speak into your microphone and get real-time transcription!",
+)
+
+if __name__ == "__main__":
+    iface.launch()