About Veena

Veena is a speech to text application built for India's languages. It turns spoken words into written text in all 22 scheduled languages of the Indian Constitution — from Hindi, Tamil, Telugu, Bengali and Marathi to Bodo, Santali, Konkani and Sanskrit.

Speech recognition for Indian languages has historically meant sending your voice to a cloud server. That is a poor fit for sensitive conversations, for people with unreliable internet, and for languages that big cloud providers ignore. Veena takes the opposite approach: everything runs on your own device.

Tap the microphone, speak in your language, and Veena converts your words into editable text you can copy or share. The recognition model itself is bundled with the app — after installation, Veena works completely offline.

Every language gets a dedicated model

Veena is built on a multilingual Conformer model trained on all 22 languages. Each language has its own vocabulary and decoding configuration, so you are not relying on a generic "one model for everything" approximation.

On the desktop, you can choose between models that trade accuracy for speed, download additional or custom models, and delete them when you are done.

Private and secure by design

No account. No cloud. No telemetry. Your recordings and transcripts are stored locally and are never uploaded anywhere. Veena only asks for the microphone permission, and nothing else.

On every platform you own

Veena is available as an app for Android and iOS, and as a desktop application for Windows, macOS and Linux. Whatever device you use, your speech stays on that device.

Made by an independent developer

Veena is developed by Nikhil Reddy Avuthu. It is a small, independent project — feedback, feature requests, and bug reports are welcome at anr.appcontact@gmail.com.

Your language deserves speech recognition that respects it — and respects your privacy. That is what Veena is for.