	In Browser
	StumbleUpon
	del.icio.us
	Google
	Google Buzz
	reddit
	LinkedIn

	Facebook
	Twitter
	Linkedin
	E-Mail

Generative AI > Google Gemini API > Gemini API for Language Translation and Localisation

Gemini API for Language Translation and Localisation

Author: Venkata Sudhakar

India is a multilingual market. Reaching customers in Hindi, Tamil, Telugu, Kannada, Malayalam, Bengali, Marathi, and Gujarati is essential for true nationwide reach. The Gemini API translates not just words but cultural context - it understands that product descriptions need different tone and emphasis for different regional markets, and that customer service responses need natural-sounding language rather than literal translations.

ShopMax India launched regional language product pages to reach Tier 2 and Tier 3 city customers who prefer browsing in their native language. Using Gemini, the content team translates and localises product descriptions, category pages, and customer notifications into 8 Indian languages. Conversion rates on regional language pages are 34% higher than English pages in the same markets, validating the investment.

The below example shows how to translate product descriptions into multiple Indian languages with cultural localisation, not just literal translation.

import vertexai
from vertexai.generative_models import GenerativeModel
import json

vertexai.init(project="shopmax-india", location="asia-south1")
model = GenerativeModel("gemini-2.0-flash")

SUPPORTED_LANGUAGES = {
    "hi": "Hindi",
    "ta": "Tamil",
    "te": "Telugu",
    "kn": "Kannada",
    "ml": "Malayalam",
    "bn": "Bengali",
    "mr": "Marathi",
    "gu": "Gujarati"
}

def translate_product_content(content: str, target_lang_code: str,
                               content_type: str = "product_description") -> dict:
    lang_name = SUPPORTED_LANGUAGES.get(target_lang_code, target_lang_code)

prompt = f"""Translate and localise this {content_type} for ShopMax India.
Target language: {lang_name} ({target_lang_code})

Instructions:
- Translate naturally, not literally - use language a native speaker would use
- Keep product names, model numbers, and prices in their original form
- Adapt tone for the regional market (conversational and warm for South Indian languages,
  direct and clear for North Indian languages)
- Keep the same meaning and key selling points

Return ONLY a JSON object with:
- translated_text: the translated content
- language: "{lang_name}"
- back_translation: brief English back-translation to verify accuracy

Content to translate:
{content}

JSON:
"""

response = model.generate_content(
        prompt,
        generation_config={"temperature": 0.3}
    )
    return json.loads(response.text)

product_desc = (
    "Sony WH-1000XM5 Wireless Headphones - Industry-leading noise cancellation "
    "with 30-hour battery life. Perfect for work-from-home and travel. "
    "Available at ShopMax India for Rs 34,990 with free delivery."
)

for lang_code in ["hi", "ta", "te"]:
    result = translate_product_content(product_desc, lang_code)
    print(f"Language: {result['language']}")
    print(f"Translation: {result['translated_text'][:100]}...")
    print(f"Verified: {result['back_translation'][:80]}...")
    print()

It gives the following output,

Language: Hindi
Translation: Sony WH-1000XM5 वायरलेस हेडफोन - बेहतरीन नॉइज़ कैंसलेशन के साथ...
Verified: Sony WH-1000XM5 wireless headphones with excellent noise cancellation...

Language: Tamil
Translation: Sony WH-1000XM5 வயர்லெஸ் ஹெட்ஃபோன் - தொழில்துறை முன்னணி...
Verified: Sony WH-1000XM5 wireless headphones - industry-leading noise cancellation...

Language: Telugu
Translation: Sony WH-1000XM5 వైర్లెస్ హెడ్ఫోన్స్ - అత్యుత్తమ నాయిస్ క్యాన్సెలేషన్...
Verified: Sony WH-1000XM5 wireless headphones with outstanding noise cancellation...

The below example shows how to translate customer service notification templates for SMS and push notifications.

NOTIFICATION_TEMPLATES = {
    "order_shipped": (
        "Your order {order_id} has been shipped! "
        "Expected delivery: {delivery_date}. Track at shopmax.in/track"
    ),
    "order_delivered": (
        "Your order {order_id} has been delivered. "
        "We hope you love your purchase! Rate your experience on ShopMax India."
    )
}

def get_localised_notification(template_key: str, lang_code: str,
                                variables: dict) -> str:
    template = NOTIFICATION_TEMPLATES[template_key]

# Translate the template (with placeholders preserved)
    prompt = f"""Translate this SMS notification template to {SUPPORTED_LANGUAGES[lang_code]}.
IMPORTANT: Keep {{order_id}}, {{delivery_date}} placeholders exactly as-is.
Keep brand name "ShopMax India" as-is. Make it warm and natural.
Return ONLY the translated text, nothing else.

Template: {template}
"""

response = model.generate_content(
        prompt, generation_config={"temperature": 0.2}
    )
    translated = response.text.strip()

# Apply variable substitution
    for key, value in variables.items():
        translated = translated.replace("{" + key + "}", str(value))

return translated

# Generate localised order shipped SMS for Tamil customer
sms = get_localised_notification(
    "order_shipped", "ta",
    {"order_id": "ORD-58291", "delivery_date": "April 7"}
)
print(f"Tamil SMS: {sms}")

It gives the following output,

Tamil SMS: உங்கள் ஆர்டர் ORD-58291 அனுப்பப்பட்டது!
எதிர்பார்க்கப்படும் டெலிவரி: April 7.
ShopMax India இல் shopmax.in/track மூலம் கண்காணிக்கவும்.

The back-translation verification field is a quality assurance mechanism - if the back-translated English differs significantly from the original, it flags the translation for human review before going live. ShopMax India uses this pattern to maintain translation quality at scale, with automated back-translation checks catching about 3% of translations that need human correction before publishing to regional language pages.

Send your comments, suggestions or queries regarding this site to [email protected].