Voisty: how to use it and how it works inside

Potok transcribes Russian speech; this site has been translated into English for convenience.

Voisty is a dictation app that recognizes speech on the device itself, without sending audio to a server: you record a phrase and the program transcribes it locally on a Windows or Mac computer, with offline models such as Whisper and Parakeet. Only 5 minutes are free, to evaluate quality; after that a paid subscription is needed. Potok (potok.now), for comparison, recognizes speech in the cloud and accepts payment in rubles.

Voisty is a dictation and speech transcription program that does all audio processing locally, on the computer itself, and not on the developer's server. For people who are wary of cloud recognition and do not want their voice to leave the device at all, this is the main difference between the program and most of its competitors. It works on Windows and macOS and is sold by subscription or as a one-time "forever" license. Its free allowance is only 5 minutes of recognition, which is enough for a trial rather than for a working day.

How Voisty works

The key design of Voisty is offline recognition models that run directly on the user's computer: the developer names Whisper Large V3 Turbo and NVIDIA Parakeet among them. Audio does not go to a remote server at any stage: the recording, the transcription and the finished text all stay on one device. This approach means the program does not need a constant network connection to recognize speech: everything required to work is already inside the app itself, not on the developer's side. For those who work with sensitive information and cannot allow audio to leave the computer, this removes the question entirely: there is no need to check a third-party service's data storage policy, because there is no third-party service in the chain.

The offline approach has a downside. A local model runs on the user's own processor or graphics card, not on a specialized server. The developer lists more than ninety supported languages, including Russian, but there is no separate public data on the recognition quality of Russian speech specifically on the Voisty site, at least we did not come across any. The design of the program is described in more detail in the review on the developer's site.

Sources: voisty.com · voisty.com

Who it suits

• those who must rule out passing audio anywhere else: lawyers, doctors, and anyone who works with documents under a non-disclosure agreement;
• users without a stable internet connection: the offline model works without a network if the program is already installed and the model was downloaded in advance;
• those who are ready to pay once for a lifetime license and do not want to think about a subscription every month;
• those with a fairly powerful computer: offline recognition runs on it directly, without contacting a server.

It does not suit those who type mostly in Russian and want to know the recognition quality for that language in advance: this is exactly the data the developer's site lacks. A similar review of an offline option is in the article on Aqua Voice for Russian.

Limits for a user from Russia

The main difficulty is payment. The site lists Visa, Mastercard, Amex and PayPal. Whether Voisty accepts cards of Russian banks is not stated separately on the site, and there is no confirmed data on this. The developer's legal entity is not named on the site, so there is no one to address a question about paying from Russia directly.

There is a separate uncertainty about language: Russian is in the list of ninety-plus supported languages, but the developer does not describe the recognition quality for it separately. According to the site's data, it is confirmed neither by a number nor by a test. For comparison, Potok has its own model trained for Russian speech: this is a fact from the service's data, not from Voisty's documentation.

On the plus side, the dictation mechanism itself does not depend on the service's localization: since audio does not leave the device, the question of where servers are located and of personal data laws simply does not arise, unlike with cloud programs.

When Potok is not a replacement for Voisty

If it is essential for you that audio never leaves the computer, Voisty is objectively closer to the task: Potok is a cloud service, recognition runs on servers in Russia, and it does not work at all without the internet. At a meeting without a network, or in a room where audio must not go outside the company perimeter, an offline program wins by definition, and changing the service cannot help here.

It also makes sense to stay with Voisty if you have already bought a "forever" license and cannot get the money back: a one-time payment pays off over time, while switching to a subscription means repeated spending. Another argument is the familiar work of offline models on your own hardware: if it is already set up and its speed suits you, there is no need to replace it with a cloud one. Other options for comparison are collected in the section comparison of voice input programs.

Frequently asked questions

Primary sources: voisty.com · voisty.com