Create app.py

app.py

@@ -0,0 +1,63 @@
+import gradio as gr
+import tempfile
+import numpy as np
+from faster_whisper import WhisperModel
+from scipy.io.wavfile import write
+
+model = WhisperModel(
+    "small",
+    device="cpu",
+    compute_type="int8"
+)
+
+last_text = ""
+
+def transcribe(audio):
+    global last_text
+
+    if audio is None:
+        return ""
+
+    sr, y = audio
+
+    with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as f:
+        write(f.name, sr, y.astype(np.int16))
+
+        segments, _ = model.transcribe(
+            f.name,
+            language="en"
+        )
+
+    text = " ".join(
+        segment.text for segment in segments
+    ).strip()
+
+    if text == last_text:
+        return last_text
+
+    last_text = text
+    return text
+
+
+with gr.Blocks() as demo:
+
+    gr.Markdown("# Real-Time English Speech Recognition")
+
+    audio = gr.Audio(
+        streaming=True,
+        sources=["microphone"],
+        type="numpy"
+    )
+
+    text = gr.Textbox(
+        label="Live Transcript",
+        lines=10
+    )
+
+    audio.stream(
+        transcribe,
+        audio,
+        text
+    )
+
+demo.launch()