Spaces:
Runtime error
Runtime error
Create app.py
Browse files
app.py
ADDED
|
@@ -0,0 +1,115 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
from transformers import pipeline
|
| 2 |
+
from langdetect import detect
|
| 3 |
+
from gtts import gTTS
|
| 4 |
+
import pyttsx3
|
| 5 |
+
import gradio as gr
|
| 6 |
+
import os
|
| 7 |
+
import tempfile
|
| 8 |
+
|
| 9 |
+
# π Supported language codes
|
| 10 |
+
LANG_CODE = {
|
| 11 |
+
"en": "English", "es": "Spanish", "fr": "French", "de": "German",
|
| 12 |
+
"it": "Italian", "nl": "Dutch", "ru": "Russian", "zh": "Chinese"
|
| 13 |
+
}
|
| 14 |
+
|
| 15 |
+
# Reverse lookup
|
| 16 |
+
LANG_NAME_TO_CODE = {v: k for k, v in LANG_CODE.items()}
|
| 17 |
+
|
| 18 |
+
# Cache for loaded models
|
| 19 |
+
translation_cache = {}
|
| 20 |
+
|
| 21 |
+
# Detect language
|
| 22 |
+
def detect_language(text):
|
| 23 |
+
try:
|
| 24 |
+
lang_code = detect(text)
|
| 25 |
+
return LANG_CODE.get(lang_code, lang_code)
|
| 26 |
+
except:
|
| 27 |
+
return "Unknown"
|
| 28 |
+
|
| 29 |
+
# Load pipeline per language pair
|
| 30 |
+
def get_translation_pipeline(src_code, tgt_code):
|
| 31 |
+
model_name = f"Helsinki-NLP/opus-mt-{src_code}-{tgt_code}"
|
| 32 |
+
key = (src_code, tgt_code)
|
| 33 |
+
if key not in translation_cache:
|
| 34 |
+
try:
|
| 35 |
+
translation_cache[key] = pipeline("translation", model=model_name, device=-1)
|
| 36 |
+
except:
|
| 37 |
+
translation_cache[key] = None
|
| 38 |
+
return translation_cache[key]
|
| 39 |
+
|
| 40 |
+
# Offline TTS
|
| 41 |
+
def speak_text(text, lang="en"):
|
| 42 |
+
engine = pyttsx3.init()
|
| 43 |
+
engine.setProperty("rate", 150)
|
| 44 |
+
engine.setProperty("voice", lang)
|
| 45 |
+
engine.say(text)
|
| 46 |
+
engine.runAndWait()
|
| 47 |
+
|
| 48 |
+
# Optional audio file (gTTS)
|
| 49 |
+
def generate_audio_file(text, lang_code):
|
| 50 |
+
try:
|
| 51 |
+
tts = gTTS(text=text, lang=lang_code)
|
| 52 |
+
temp_file = tempfile.NamedTemporaryFile(delete=False, suffix=".mp3")
|
| 53 |
+
tts.save(temp_file.name)
|
| 54 |
+
return temp_file.name
|
| 55 |
+
except:
|
| 56 |
+
return None
|
| 57 |
+
|
| 58 |
+
# Main translation logic
|
| 59 |
+
def translate_text(input_text, target_lang, speak=False):
|
| 60 |
+
if not input_text.strip():
|
| 61 |
+
return "Please enter some text to translate.", None, "Unknown"
|
| 62 |
+
|
| 63 |
+
detected_lang = detect_language(input_text)
|
| 64 |
+
if detected_lang == "Unknown":
|
| 65 |
+
return "Could not detect source language.", None, detected_lang
|
| 66 |
+
|
| 67 |
+
if detected_lang == target_lang:
|
| 68 |
+
return f"Source and target languages are the same. Text: {input_text}", None, detected_lang
|
| 69 |
+
|
| 70 |
+
src_code = LANG_NAME_TO_CODE.get(detected_lang)
|
| 71 |
+
tgt_code = LANG_NAME_TO_CODE.get(target_lang)
|
| 72 |
+
|
| 73 |
+
if not src_code or not tgt_code:
|
| 74 |
+
return f"Unsupported language pair: {detected_lang} β {target_lang}", None, detected_lang
|
| 75 |
+
|
| 76 |
+
translator = get_translation_pipeline(src_code, tgt_code)
|
| 77 |
+
if not translator:
|
| 78 |
+
return f"No model found for {detected_lang} β {target_lang}", None, detected_lang
|
| 79 |
+
|
| 80 |
+
try:
|
| 81 |
+
result = translator(input_text, max_length=512)
|
| 82 |
+
translated_text = result[0]['translation_text']
|
| 83 |
+
audio_path = generate_audio_file(translated_text, tgt_code) if speak else None
|
| 84 |
+
return translated_text, audio_path, detected_lang
|
| 85 |
+
except Exception as e:
|
| 86 |
+
return f"Translation failed: {str(e)}", None, detected_lang
|
| 87 |
+
|
| 88 |
+
# Gradio UI
|
| 89 |
+
with gr.Blocks() as app:
|
| 90 |
+
gr.Markdown("# π AI Translator with Speech & Auto-Language Detection")
|
| 91 |
+
|
| 92 |
+
with gr.Row():
|
| 93 |
+
with gr.Column():
|
| 94 |
+
input_text = gr.Textbox(lines=4, label="Enter text to translate")
|
| 95 |
+
target_lang = gr.Dropdown(
|
| 96 |
+
choices=list(LANG_NAME_TO_CODE.keys()),
|
| 97 |
+
value="French",
|
| 98 |
+
label="Target Language"
|
| 99 |
+
)
|
| 100 |
+
speak_checkbox = gr.Checkbox(label="π Enable speech output", value=True)
|
| 101 |
+
translate_button = gr.Button("π Translate")
|
| 102 |
+
|
| 103 |
+
with gr.Column():
|
| 104 |
+
output_text = gr.Textbox(lines=4, label="Translated Text")
|
| 105 |
+
output_audio = gr.Audio(label="Speech Output", autoplay=True)
|
| 106 |
+
detected_lang = gr.Textbox(label="Detected Language", interactive=False)
|
| 107 |
+
|
| 108 |
+
translate_button.click(
|
| 109 |
+
translate_text,
|
| 110 |
+
inputs=[input_text, target_lang, speak_checkbox],
|
| 111 |
+
outputs=[output_text, output_audio, detected_lang]
|
| 112 |
+
)
|
| 113 |
+
|
| 114 |
+
if __name__ == "__main__":
|
| 115 |
+
app.launch(share=True, server_name="0.0.0.0", server_port=7862)
|