Update app.py

app.py

@@ -1,70 +1,74 @@
-# Text-→Translation-→Speech  |  CPU-only Hugging-Face Space
+# Text → Translate → Speech   |   CPU-only Hugging-Face Space
 
 import tempfile
 from pathlib import Path
 
-import gradio as gr
-from transformers import pipeline, AutoTokenizer, AutoModelForSeq2SeqLM
+import torch
+import TTS.utils.radam as tts_radam
+from transformers import pipeline
 from TTS.api import TTS
+import gradio as gr
 
 # ─────────────────────────────
-# Translation pipelines
+# 1. Allow-list Coqui’s custom RAdam class (Torch ≥2.6)
 # ─────────────────────────────
-PIPE_EN_ES = pipeline(
-    "translation", model="Helsinki-NLP/opus-mt-en-es", device=-1
-)
-PIPE_ES_EN = pipeline(
-    "translation", model="Helsinki-NLP/opus-mt-es-en", device=-1
-)
+torch.serialization.add_safe_globals({"TTS.utils.radam.RAdam": tts_radam.RAdam})
 
 # ─────────────────────────────
-# TTS models (Coqui TTS)
+# 2. Translation pipelines
+# ─────────────────────────────
+PIPE_EN_ES = pipeline("translation", model="Helsinki-NLP/opus-mt-en-es", device=-1)
+PIPE_ES_EN = pipeline("translation", model="Helsinki-NLP/opus-mt-es-en", device=-1)
+
+# ─────────────────────────────
+# 3. TTS models (Coqui TTS)
 # ─────────────────────────────
 TTS_EN = TTS(model_name="tts_models/en/ljspeech/tacotron2-DDC", progress_bar=False)
 TTS_ES = TTS(model_name="tts_models/es/mai/tacotron2-DDC", progress_bar=False)
 
 # ─────────────────────────────
-# Core function
+# 4. Helper: synthesize WAV to temp file
+# ─────────────────────────────
+def synthesize(tts_model: TTS, text: str) -> str:
+    with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as tmp:
+        wav_path = Path(tmp.name)
+    tts_model.tts_to_file(text=text, file_path=wav_path)
+    return str(wav_path)
+
+# ─────────────────────────────
+# 5. Core translate-and-speak function
 # ─────────────────────────────
-def translate_and_speak(text: str, target_lang: str):
-    if not text.strip():
+def translate_and_speak(txt: str, target_lang: str):
+    if not txt.strip():
         return "", None
 
     if target_lang == "Spanish":
-        translated = PIPE_EN_ES(text)[0]["translation_text"]
-        wav_path = synthesize(TTS_ES, translated)
+        translated = PIPE_EN_ES(txt)[0]["translation_text"]
+        audio_file = synthesize(TTS_ES, translated)
     else:  # English
-        translated = PIPE_ES_EN(text)[0]["translation_text"]
-        wav_path = synthesize(TTS_EN, translated)
-
-    return translated, wav_path
+        translated = PIPE_ES_EN(txt)[0]["translation_text"]
+        audio_file = synthesize(TTS_EN, translated)
 
-
-def synthesize(tts_model: TTS, text: str) -> str:
-    """Generate WAV and return file path."""
-    with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as tmp:
-        wav_file = Path(tmp.name)
-    tts_model.tts_to_file(text=text, file_path=wav_file)
-    return str(wav_file)
+    return translated, audio_file
 
 # ─────────────────────────────
-# Gradio UI
+# 6. Gradio UI
 # ─────────────────────────────
 with gr.Blocks(title="Translator & TTS") as demo:
     gr.Markdown(
         "# 🌐💬 Text → Translate → Speech\n"
-        "Type a sentence, choose target language, hear it spoken!"
+        "Type a sentence, choose target language, and hear it spoken."
     )
 
-    inp = gr.Textbox(label="Your sentence (English or Spanish)", lines=2)
-    tgt = gr.Radio(["Spanish", "English"], value="Spanish", label="Translate to")
+    text_in = gr.Textbox(label="Sentence (English or Spanish)", lines=2)
+    lang    = gr.Radio(["Spanish", "English"], value="Spanish", label="Translate to")
 
-    btn = gr.Button("Translate & Speak", variant="primary")
+    run_btn = gr.Button("Translate & Speak", variant="primary")
 
-    out_text  = gr.Textbox(label="Translated text", interactive=False)
-    out_audio = gr.Audio(label="TTS output", type="filepath")
+    text_out = gr.Textbox(label="Translated text", interactive=False)
+    wav_out  = gr.Audio(label="Speech output", type="filepath")
 
-    btn.click(translate_and_speak, [inp, tgt], [out_text, out_audio])
+    run_btn.click(translate_and_speak, [text_in, lang], [text_out, wav_out])
 
 if __name__ == "__main__":
     demo.launch(server_name="0.0.0.0")