Air-Gapped & Offline Inference

Offline AI Models

Empower IndiCAT with complete on-device privacy. Download our curated neural translation and LLM packages to translate documents, subtitles, and scripts without an internet connection.

Indian Languages Suite

IndicTrans2

AI4Bharat's state-of-the-art Sequence-to-Sequence neural machine translation model. Tuned specifically for all 22 official scheduled languages of India with cultural accuracy, honorific nuances, and transliteration awareness.

Engine Architecture Transformer NMT
Package Format ZIP Archive (.zip)
Estimated Size ~1.2 GB
Execution Mode CPU & CUDA (Direct)
Major Languages Supported
Hindi Telugu Tamil Kannada Malayalam Bengali Marathi Gujarati Punjabi Odia Assamese Urdu + 10 More
Download IndicTrans2 (.zip)
Asian & Regional Languages

Qwen Asian Engine

High-accuracy dense multilingual model fine-tuned for East Asian, Southeast Asian, and Middle-Eastern linguistic structures. Handles CJK character tokenization and tonal syntax with unmatched fidelity.

Engine Architecture Multilingual LLM
Package Format ZIP Archive (.zip)
Estimated Size ~2.4 GB
Execution Mode Quantized GGUF / ONNX
Major Languages Supported
Chinese (Simp/Trad) Japanese Korean Vietnamese Thai Indonesian Malay Tagalog Arabic
Download Qwen Model (.zip)
Global & World Languages

NLLB (Meta AI)

Meta AI's No Language Left Behind universal translation engine. Delivers comprehensive translation across 200+ global languages, spanning European, African, Latin American, and Indo-European language pairs.

Engine Architecture Universal NMT (200-L)
Package Format ZIP Archive (.zip)
Estimated Size ~1.8 GB
Execution Mode Local Inference
Major Languages Supported
Spanish French German Portuguese Russian Italian Dutch Turkish Swahili + 190 More
Download NLLB Model (.zip)

How to Install Offline Models

Follow these 3 simple steps to enable air-gapped translation inside your IndiCAT Desktop app.

1

Download Package

Click the download button above for your target model pack. Save the downloaded .zip archive to your computer.

2

Extract to Assets

Extract the contents of the ZIP archive directly into your IndiCAT models directory:

IndiCAT\Assets\ai_models\
3

Launch & Translate

Open IndiCAT. In Settings ➔ Engine Mode, switch to Offline Local Engine. IndiCAT detects the models automatically!