# Translation API for Indian Languages (Mayura)

> Translation that feels native across India's languages.

Translate text between English and 10 Indian languages with accuracy that preserves meaning, tone, and cultural context. Mayura is Sarvam's translation model, purpose-built for Indian languages. It translates directly between Indian languages and English without routing through a bridge language, preserving meaning, tone, and cultural context.

---

## Supported Languages

Mayura supports translation between English and 10 Indian languages -- 11 languages total, covering all major language families across India.

| Language | Native Script | Code |
|----------|--------------|------|
| Hindi | (hi-IN) | hi-IN |
| Tamil | (ta-IN) | ta-IN |
| Bengali | (bn-IN) | bn-IN |
| Telugu | (te-IN) | te-IN |
| Kannada | (kn-IN) | kn-IN |
| Malayalam | (ml-IN) | ml-IN |
| Marathi | (mr-IN) | mr-IN |
| Gujarati | (gu-IN) | gu-IN |
| Punjabi | (pa-IN) | pa-IN |
| Odia | (or-IN) | or-IN |
| English | (en-IN) | en-IN |

Translation works in all directions between these languages, not just to and from English.

---

## Why Translate with Sarvam

### Direct Translation Between Indian Languages
Other systems use English as a bridge. Mayura translates directly between Indian languages, reducing the distortion that accumulates every time meaning passes through an intermediary.

### Long-Context Understanding
Paragraph-level and document-level translation that maintains consistency of terminology, tone, and narrative across the full text, not just sentence by sentence.

### Code-Mixed Input, Handled Naturally
Users write in Hindi-English, Tamil-English, or any mix they are comfortable with. Mayura reads it as-is and translates it faithfully without requiring clean single-language input.

### Numeral and Script Flexibility
Switch between Indic and international numeral formats. Choose native script or Roman transliteration for the output. Set it once in the API call.

### Automatic Language Detection
Set source_language_code to "auto" and let the model detect the source language automatically. No need to know the input language in advance.

---

## Translation Modes

Mayura offers four translation modes so the output matches where and how it will be read.

### Formal
Professional, respectful tone for official documents, legal content, government communications, and enterprise workflows. Use this for anything that requires an official register.

### Modern Colloquial
Casual, contemporary language used in daily conversation. Best for consumer apps, chat interfaces, and social contexts. Reflects how people actually speak today.

### Classical Colloquial
Traditional everyday speech with cultural nuances preserved. Closer to how an older generation would speak and write. Preserves heritage and traditional phrasing.

### Code-Mixed
Handles mixed-language input from Indian messaging, social media, and customer conversations with output that mirrors the blend. Input like "top-notch quality thi, delivery thodi late thi" stays as it is.

---

## How to Translate with Sarvam

A four-step API workflow from plain text to fluent output.

1. **Send your source text** -- POST plain text in any of the 11 supported languages. Code-mixed input works as-is.
2. **Set source and target languages** -- Specify language codes, or set source to "auto" for automatic detection.
3. **Choose mode and output format** -- Pick a translation mode (formal, colloquial, code-mixed) and set script preferences.
4. **Receive the translated text** -- Get fluent, formatted output via the REST API, ready to display or pipe downstream.

---

## Gender-Aware Translation

Mayura supports gender-aware translation across Indian languages, out of the box. Many Indian languages have gendered verb forms, pronouns, and adjectives that change based on the speaker or subject's gender. Mayura handles this correctly, producing translations that respect gender context. This is especially important for customer support, chatbot responses, and personalized communications.

---

## Use Cases

### Product and App Localization
Your app speaks every language your users do. Every word they see translates without anything breaking. Ship faster, in more markets.
- In-app strings and UI copy
- Push notifications and alerts
- Onboarding and help content

### EdTech and Learning
Formal mode for academic content. Natural, conversational mode for tutoring and video captions.
- Course and curriculum content
- Tutoring and live session captions
- Assessments and answer explanations

### BFSI and Fintech
Financial notifications that people actually understand. EMI reminders, KYC prompts, and loan updates in the language they think in, the numerals they use, the names they know. Ready to send.
- EMI reminders and KYC prompts
- Loan updates and account summaries
- Native numeral and script formatting

### E-Commerce Personalization
Translate product reviews, promo messages, and order updates with code-mixing intact.
- Product reviews with code-mixing intact
- Promo messages and offers
- Order and delivery updates

### Customer Support
Support customers in the language they think in, not just the language they type in. Gender-aware translation across Indian languages, out of the box.
- Gender-aware agent responses
- Chatbot and IVR localisation
- Escalation scripts across languages

---

## Pricing

Simple, transparent pricing. Start free. Scale as you grow. No hidden costs.

| Model | Price | Unit |
|-------|-------|------|
| Sarvam Translate V1 | Rs 20 | per 10K characters |
| Translate Mayura V1 | Rs 20 | per 10K characters |

**What is included:**
- Free trial included
- No credit card required
- Get API keys instantly
- 11 languages supported
- 4 translation modes
- Code-mixed input support
- Automatic language detection
- Volume discounts available
- Enterprise pricing available

---

## API Details

- **Method:** POST
- **Input:** Plain text in any of the 11 supported languages; code-mixed input accepted
- **Output:** Translated text via REST API
- **Language detection:** Set source_language_code to "auto" for automatic detection
- **Numeral format:** Configurable -- native Indic or international digits
- **Script preference:** Configurable -- native script or Roman transliteration
- **Model name:** Mayura

---

## FAQ

**What is Mayura?**
Mayura is Sarvam's translation model, purpose-built for Indian languages. It translates directly between Indian languages and English without routing through a bridge language, preserving meaning, tone, and cultural context.

**Which languages does Mayura support?**
Mayura supports translation between English and 10 Indian languages: Hindi, Tamil, Bengali, Telugu, Kannada, Malayalam, Marathi, Gujarati, Punjabi, and Odia.

**What are the translation modes?**
Mayura offers four modes: Formal (professional, official tone), Modern colloquial (casual, contemporary), Classical colloquial (traditional everyday speech), and Code-mixed (handles mixed-language input and produces natural blended output).

**Can Mayura handle code-mixed input?**
Yes. Mayura reads Hindi-English, Tamil-English, and other mixed-language text as-is and translates it faithfully without requiring clean single-language input.

**Does Mayura support automatic language detection?**
Yes. Set source_language_code to "auto" in your API call and Mayura will detect the source language automatically.

**Can I control numeral and script format in the output?**
Yes. You can configure numeral format (native Indic or international digits) and script preference (native script or Roman transliteration) directly in your API call.
