QVAC by Tether

The QVAC SDK puts on-device AI inside your own app.

One install covers 12 AI tasks through a single API, including text generation, speech to text, text to speech, translation, image and video generation, image understanding, OCR, embeddings, search over your own documents, and fine-tuning a model on the device itself.

It runs on iOS, Android, macOS, Linux and Windows, in JavaScript or TypeScript. There is an official Python SDK on PyPI at the same version. Models load from a file on disk or straight from Hugging Face.

Because the work happens on the device, there is no inference bill: ten thousand installs cost you the same as ten. Your users' data stays on their own device, and the app keeps working with no connection at all.

github.com/tetherto/qvac

docs.qvac.tether.io

QVAC is open-source AI that runs on the device in front of you, with no cloud account and no API bill.

We build it in the open. The SDK is Apache 2.0, the model weights are published on Hugging Face, and the training datasets are open for research and education.

Cloud AI needs a credit card, a steady connection and a currency that is easy to spend online. We want intelligence to be a universal right rather than a subscription: free to run on a device someone already owns, in the language they speak at home.

QVAC is built by Tether.

qvac.tether.io