README.md

metadata

title: Sarvam Translate
emoji: 🌐
colorFrom: purple
colorTo: blue
sdk: gradio
sdk_version: 5.35.0
app_file: app.py
pinned: false
short_description: Sarvam-Translate for 23 Indian Languages

Sarvam Translation Tool

A translation application using the Sarvam-Translate model (Gemma3-4B) for translating between 23 Indian languages. Supports direct translation between any pair of languages without English as an intermediate step.

Features

• Text Translation: Direct text input and output
• Document Translation: Upload PDF/DOCX files and download translated documents
• 23 Language Support: All scheduled Indian languages plus English
• Direct Translation: Any language pair translated directly (no pivoting through English)
• Long Context: 8K token context window for better coherence across paragraphs

Supported Languages

Assamese, Bengali, Bodo, Dogri, English, Gujarati, Hindi, Kannada, Kashmiri, Konkani, Maithili, Malayalam, Manipuri, Marathi, Nepali, Odia, Punjabi, Sanskrit, Santali, Sindhi, Tamil, Telugu, Urdu

Usage

Text Translation

1. Select the "Text Translation" tab
2. Enter or paste your text in the input box
3. Choose source and target languages
4. Click "Translate Text"
5. View the translated text in the output box

Document Translation

1. Select the "Document Translation" tab
2. Upload a PDF or DOCX file
3. Choose source and target languages
4. Click "Translate Document"
5. Download the translated document when ready

Technical Details

• Model: sarvamai/sarvam-translate (Gemma3-4B-IT based)
• Framework: Transformers, PyTorch
• Interface: Gradio
• Supported File Types: PDF, DOCX
• Context Window: 8,192 tokens
• GPU: NVIDIA L4 (24GB VRAM)

License

GPL-3.0, following the Sarvam-Translate model license.