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

	Facebook
	Twitter
	Linkedin
	E-Mail

Generative AI > Google Gemini API > Gemini API Translation and Multilingual Support

Gemini API Translation and Multilingual Support

Author: Venkata Sudhakar

India has 22 official languages and hundreds of regional dialects. An e-commerce platform that supports customers in Hindi, Tamil, Telugu, Kannada, and Bengali has a significant advantage. Gemini handles Indian language translation with high accuracy, understanding colloquial phrasing and regional context that generic translation APIs miss. ShopMax India uses Gemini to respond to customer queries in the language the customer wrote in.

The workflow: detect the language of the incoming message, translate it to English for agent processing, get the response, then translate back to the customer language. This keeps your agents English-only while serving customers in any language.

The below example shows how ShopMax India implements auto-language detection and multilingual customer support.

import google.generativeai as genai

genai.configure(api_key="your-api-key")
model = genai.GenerativeModel("gemini-2.0-flash")

def detect_and_translate(text: str, target_lang: str = "English") -> dict:
    response = model.generate_content(
        f"""Detect the language and translate to {target_lang}.
Return JSON: {{"detected_language": "...", "translation": "..."}}
Text: {text}""",
        generation_config=genai.GenerationConfig(temperature=0.0)
    )
    import json
    return json.loads(response.text.strip().strip("```json").strip("```"))

def translate_response(english_response: str, target_language: str) -> str:
    if target_language.lower() == "english":
        return english_response
    response = model.generate_content(
        f"Translate this customer support response to {target_language}. "
        f"Keep product names, prices (Rs amounts), and order IDs unchanged.\n\n{english_response}",
        generation_config=genai.GenerationConfig(temperature=0.1)
    )
    return response.text.strip()

# Test with different Indian languages
messages = [
    "Mera order kab aayega? Order number ORD-2024-88712",  # Hindi
    "En order eppodhu varum?",  # Tamil
    "Nanna order yavaaga baruttade?",  # Kannada
]
for msg in messages:
    result = detect_and_translate(msg)
    print(f"Original: {msg}")
    print(f"Detected: {result['detected_language']}")
    print(f"English: {result['translation']}")
    print()

It gives the following output,

Original: Mera order kab aayega? Order number ORD-2024-88712
Detected: Hindi
English: When will my order arrive? Order number ORD-2024-88712

Original: En order eppodhu varum?
Detected: Tamil
English: When will my order come?

Original: Nanna order yavaaga baruttade?
Detected: Kannada
English: When will my order arrive?

Full multilingual support pipeline for ShopMax India customer service,

It gives the following output,

[Hindi] Query: Mujhe apna order wapas karna hai, product damage tha
[Hindi] Response: Namaste! Hum aapke damaged product ke liye maafi chahte hain.
ShopMax India ki return policy ke anusar, aap 7 din ke andar return kar sakte hain.
Apna order number share karein aur hum pickup arrange karenge.

For ShopMax India, prioritise Hindi, Tamil, Telugu, Kannada, Bengali, and Marathi as these cover 80% of the Indian population. Store language preference per customer ID in Firestore so returning customers always get responses in their preferred language without re-detection.

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