Update app.py

app.py

@@ -1,19 +1,17 @@
 # =============================================================
-# Hugging Face Space – Lecture → Multilingual Podcast Generator
+# Hugging Face Space – Lecture → Podcast Generator (User‑selectable Languages)
 # =============================================================
-# * **Text generation** – SmolAgents `HfApiModel` running the remote
-#   Qwen/Qwen2.5‑Coder‑32B‑Instruct model.
-# * **Speech synthesis** – `huggingface_hub.InferenceClient.text_to_speech`
-#   (serverless) with open models per language – no heavy local
-#   downloads.
-# * Outputs five FLAC files (English, Bangla, Chinese, Urdu, Nepali).
+# * **Text generation** – SmolAgents `HfApiModel` (Qwen/Qwen2.5‑Coder‑32B‑Instruct).
+# * **Speech synthesis** – `huggingface_hub.InferenceClient.text_to_speech`.
+# * Users pick which languages to generate (English, Bangla, Chinese,
+#   Urdu, Nepali). Unselected languages are skipped.
 # -----------------------------------------------------------------
 
 import os
 import tempfile
 import textwrap
 from pathlib import Path
-from typing import List, Dict
+from typing import List, Dict, Tuple, Optional
 
 import gradio as gr
 from huggingface_hub import InferenceClient
@@ -41,12 +39,14 @@ client = InferenceClient(token=os.getenv("HF_TOKEN", None))
 LANG_INFO: Dict[str, Dict[str, str]] = {
     "en": {"name": "English", "tts_model": "facebook/mms-tts-eng"},
     "bn": {"name": "Bangla",  "tts_model": "facebook/mms-tts-ben"},
-    # MMS lacks mainstream Mandarin — fallback to an open Chinese TTS
     "zh": {"name": "Chinese", "tts_model": "myshell-ai/MeloTTS-Chinese"},
     "ur": {"name": "Urdu",    "tts_model": "facebook/mms-tts-urd-script_arabic"},
     "ne": {"name": "Nepali",  "tts_model": "facebook/mms-tts-npi"},
 }
 
+# Helper map: name ➜ code
+LANG_CODE_BY_NAME = {info["name"]: code for code, info in LANG_INFO.items()}
+
 PROMPT_TEMPLATE = textwrap.dedent(
     """
     You are producing a lively two‑host educational podcast in {lang_name}.
@@ -77,14 +77,25 @@ def truncate_text(text: str, limit: int = TOKEN_LIMIT) -> str:
 # Main pipeline
 # ------------------------------------------------------------------
 
-def generate_podcast(pdf: gr.File) -> List[gr.Audio]:
-    """Generate multilingual podcast from a lecture PDF."""
+def generate_podcast(pdf: gr.File, selected_lang_names: List[str]) -> List[Optional[Tuple[str, None]]]:
+    """Generate podcast audio files for chosen languages. Returns a list
+    aligned with LANG_INFO order; unselected languages yield None."""
+    # Ensure at least one language selected
+    if not selected_lang_names:
+        return [None] * len(LANG_INFO)
+
+    selected_codes = [LANG_CODE_BY_NAME[name] for name in selected_lang_names]
+
     with tempfile.TemporaryDirectory() as tmpdir:
         raw_text = extract_pdf_text(pdf.name)
         lecture_text = truncate_text(raw_text)
-        outputs: List[tuple] = []
+        outputs: List[Optional[Tuple[str, None]]] = []
 
         for code, info in LANG_INFO.items():
+            if code not in selected_codes:
+                outputs.append(None)
+                continue
+
             # 1️⃣ Draft dialogue in the target language
             prompt = PROMPT_TEMPLATE.format(lang_name=info["name"], content=lecture_text)
             dialogue: str = llm(prompt)
@@ -102,21 +113,31 @@ def generate_podcast(pdf: gr.File) -> List[gr.Audio]:
 # Gradio interface
 # ------------------------------------------------------------------
 
+language_choices = [info["name"] for info in LANG_INFO.values()]
+
+inputs = [
+    gr.File(label="Upload Lecture PDF", file_types=[".pdf"]),
+    gr.CheckboxGroup(
+        choices=language_choices,
+        value=["English"],
+        label="Select podcast language(s) to generate",
+    ),
+]
+
 audio_components = [
-    gr.Audio(label=f"{info['name']} Podcast", type="filepath")
-    for info in LANG_INFO.values()
+    gr.Audio(label=f"{info['name']} Podcast", type="filepath") for info in LANG_INFO.values()
 ]
 
 iface = gr.Interface(
     fn=generate_podcast,
-    inputs=gr.File(label="Upload Lecture PDF", file_types=[".pdf"]),
+    inputs=inputs,
     outputs=audio_components,
-    title="Lecture → Multilingual Podcast Generator",
+    title="Lecture → Podcast Generator (Choose Languages)",
     description=(
-        "Upload a lecture PDF and receive a two‑host audio podcast in five "
-        "languages (English, Bangla, Chinese, Urdu, Nepali). Dialogue is "
-        "crafted by Qwen‑32B; speech is synthesized on‑the‑fly using the "
-        "Hugging Face Inference API — no heavy downloads or GPUs required."
+        "Upload a lecture PDF, choose your desired languages, and receive a "
+        "two‑host audio podcast. Dialogue is crafted by Qwen‑32B; speech is "
+        "synthesized on‑the‑fly using the Hugging Face Inference API — "
+        "no heavy downloads or GPUs required."
     ),
 )