Update app.py

app.py

@@ -2,61 +2,61 @@ import gradio as gr
 from transformers import pipeline
 from huggingface_hub import list_models
 import langcodes
-import re
+import difflib
 
-# Tạo mapping tên ngôn ngữ → mã ISO
+# Tạo bản đồ ngôn ngữ → mã ISO
 lang_map = {}
+models = list_models(filter="facebook/mms-tts", full=True)
 
-print("🔄 Fetching available models...")
-all_models = list_models(filter="facebook/mms-tts", full=True)
-
-for model in all_models:
-    repo = model.modelId
-    code = repo.split("-")[-1].lower()
-
-    name = ""
+for model in models:
+    repo_id = model.modelId  # ví dụ: facebook/mms-tts-tur
+    code = repo_id.split("-")[-1].lower()
+    
+    # Ưu tiên lấy tên từ metadata
     if model.cardData and "language" in model.cardData:
-        name = model.cardData["language"]
+        lang_name = model.cardData["language"]
     else:
         try:
-            name = langcodes.Language.make(language=code).display_name()
+            lang_name = langcodes.Language.make(code).display_name()
         except:
-            name = code
+            lang_name = code
 
-    name_clean = re.sub(r"\s+", " ", name.strip().lower())
-    lang_map[name_clean] = code
+    lang_map[lang_name.strip().lower()] = code
 
-print(f"✅ Total languages found: {len(lang_map)}")
-print("📋 Sample languages:")
-for i, (k, v) in enumerate(lang_map.items()):
-    if i >= 30:
-        break
-    print(f" - {k} → {v}")
+def tts_fn(text, lang_input):
+    lang_key = lang_input.strip().lower()
+    code = lang_map.get(lang_key)
 
-def tts_fn(text, language):
-    lang_input = re.sub(r"\s+", " ", language.strip().lower())
-    code = lang_map.get(lang_input)
     if not code:
-        return f"❌ Unsupported language name: '{language}'", None
+        # Gợi ý
+        suggestions = difflib.get_close_matches(lang_key, lang_map.keys(), n=5)
+        msg = f"❌ Unsupported language name: '{lang_input}'"
+        if suggestions:
+            msg += f"\n🔍 Did you mean: {', '.join(suggestions)}?"
+        return msg, None
+
     model_id = f"facebook/mms-tts-{code}"
     try:
         tts = pipeline("text-to-speech", model=model_id)
         out = tts(text)
         return f"✅ Model: {model_id}", (out["audio"], "output.wav")
     except Exception as e:
-        return f"❌ Error: {str(e)}", None
+        return f"❌ Pipeline error: {str(e)}", None
 
 with gr.Blocks() as demo:
-    gr.Markdown("# 🌐 MMS-TTS: 1100+ Languages (Flexible Name Input)")
+    gr.Markdown("## 🌍 MMS-TTS: Speak over 1100 Languages\n(Just type the name, no ISO code needed!)")
+
     with gr.Row():
-        language = gr.Textbox(label="Enter language name (e.g., English, Meadow Mari)")
-        text = gr.Textbox(label="Enter your text")
+        lang_box = gr.Textbox(label="Language (e.g., English, Meadow Mari)")
+        text_box = gr.Textbox(label="Text")
+
     with gr.Row():
-        btn = gr.Button("Speak")
+        btn = gr.Button("🔊 Speak")
+
     with gr.Row():
-        out_text = gr.Textbox(label="Info")
+        out_msg = gr.Textbox(label="Output Message")
         out_audio = gr.Audio(label="Audio", type="filepath")
 
-    btn.click(fn=tts_fn, inputs=[text, language], outputs=[out_text, out_audio])
+    btn.click(fn=tts_fn, inputs=[text_box, lang_box], outputs=[out_msg, out_audio])
 
 demo.launch()