Create app.py

app.py

@@ -0,0 +1,35 @@
+import gradio as gr
+import whisper
+from transformers import MarianMTModel, MarianTokenizer
+
+# Load models once (so startup is slower, but inference is faster)
+whisper_model = whisper.load_model("small")  # you can switch to "base" or "large"
+translator_model = MarianMTModel.from_pretrained("Helsinki-NLP/opus-mt-zh-en")
+translator_tokenizer = MarianTokenizer.from_pretrained("Helsinki-NLP/opus-mt-zh-en")
+
+def transcribe_and_translate(audio):
+    # Speech → Chinese text
+    result = whisper_model.transcribe(audio, language="zh")
+    chinese_text = result["text"].strip()
+
+    # Chinese → English translation
+    inputs = translator_tokenizer(chinese_text, return_tensors="pt", padding=True)
+    translated = translator_model.generate(**inputs, max_length=512)
+    english_text = translator_tokenizer.decode(translated[0], skip_special_tokens=True)
+
+    return chinese_text, english_text
+
+# Gradio UI
+app = gr.Interface(
+    fn=transcribe_and_translate,
+    inputs=gr.Audio(sources=["microphone", "upload"], type="filepath", label="🎙️ Speak Chinese or Upload Audio"),
+    outputs=[
+        gr.Textbox(label="🈶 Chinese Text"),
+        gr.Textbox(label="🌍 English Translation")
+    ],
+    title="🎤 Chinese Voice → English Text Translator",
+    description="Speak or upload Chinese audio. It will transcribe into Chinese text and translate into English."
+)
+
+if __name__ == "__main__":
+    app.launch()