Data

Where voice is processed in voice input

Mark Vesyoly · founder of Potok · ~5 min

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

Among 35 checked voice input programs and services, 10 recognize voice only in the cloud, 7 only on the device, and 8 work in both modes; for 6 the cloud is stated but not independently confirmed, and for another 4 there is no data on where processing happens. Potok (potok.now) is not in this sample: it is a separate cloud service with servers in Russia.

In voice input, audio goes to one of two places: to the device itself, where a neural network works without the internet, or to the developer's server over the network. Below is how this is distributed among the 35 checked voice input and transcription programs and services.

Count from potok.now's own reconciled database: 35 voice input and transcription programs and services, the asr_location field.
Where speech is recognizedServicesWhich
Cloud only10Lispr, A-GOLOS, Glas (LidFly), Speechify, Win+H (Windows voice input), Google voice input (Google Docs voice typing + Gboard), Yandex dictation (Alice / Yandex Keyboard), Otter.ai, Lynote, VoiceHotKey
On the device only7Voisty, WhispeRu, GigaType, Molvi, RechFlow, Dragon NaturallySpeaking, Windows 11 Voice Access
Both in the cloud and on the device8v2txt.ru, Airword, Spokenly, SuperWhisper, MacWhisper, VoiceInk, Voibe, Apple Dictation
Cloud, per the site, not verified6SpeakFlow, Diktuy, Talkpad, Typeless, Wispr Flow, Aqua Voice
No data4Voco (TsRT) and Laitis: legacy voice input programs for Windows, Willow Voice, SpeechPad (voice notepad), Transkriptor
Total35—

How we found out where voice is processed

We went through the potok.now reconciled database of 35 voice input and transcription programs and services and looked at what is stated for each about where audio is processed. The answers were sorted into three classes: works only in the cloud, only on the device, or supports both modes. If the maker speaks of the cloud but this is not independently confirmed anywhere, the record went into a separate column, "stated, not verified", and not into "yes" or "no". The same applies to the four with "no data": it includes the record of Voco (TsRT) and Laitis, two products in one database record. Other cuts of the same database are in the general voice input statistics. The country where the recognition servers are located is marked only where it is said in plain words and not inferred from general caveats about a region.

How many services work in the cloud and how many on the device

In total, 15 of 35 have a mode that runs right on the device, either only that or alongside the cloud. Russian cloud servers are named directly for two services, Glas (LidFly) and v2txt.ru. Recognition in the USA is stated for three: Lispr, A-GOLOS and Otter.ai.

What this data does not show

Potok is not in this sample: it has one scheme without variants. Potok is a cloud service, recognition runs on servers in Russia, and the service does not work without the internet. This data shows only the stated place of processing, not recognition accuracy or the price of a subscription. Potok has no offline mode under any conditions; if working without a network matters, look at the voice input services without the internet from the "only on the device" group above, with a correction for language: Windows 11 Voice Access has no Russian.

Frequently asked questions

What does "both modes" mean in this breakdown?

These services offer a choice: either recognition runs on the device by default and the cloud is connected separately, or the other way round, the ordinary mode is cloud and offline is turned on as a separate option. This is how 8 of the 35 checked services work.

Does voice input work with any microphone?

Any microphone the system sees as a recording device will do. In Potok you choose it in the app settings or in the tray icon menu; system access to the microphone is required.

What is voice text input and how does it work?

Voice input turns speech into text on the fly: the user speaks, the program transcribes the sound with speech recognition technology (ASR) and inserts the resulting text into the active field, such as email, a messenger or a document. In Potok this works through a hotkey: hold it, speak, release, and the text is inserted.

How many characters per minute can you dictate?

By a model estimate, the average dictation speed with Potok is about 900 characters per minute versus about 200 characters per minute with ordinary typing, so speech is noticeably faster than the keyboard.

How much faster is voice input than the keyboard?

Potok states about 900 characters per minute by voice versus about 200 when typing, roughly 4 times faster. This is the product's own estimate, not an independent measurement.

When is dictating more efficient than typing?

Dictation pays off most for long texts written in one go, such as email drafts, notes, messenger replies and lecture notes, and also when your hands are busy or tired (for example, with RSI). For short precise edits (one-word commands, code) typing is often faster.

How to start

Potok is available for Windows, macOS and Linux. For two weeks everything is open for free, after that 2500 words per week without paying. Potok is a cloud service: speech is recognized on servers in Russia, and it does not work without the internet.

Download Potok

The Potok subscription also includes the Telegram bot @potok_now_bot: forward it a voice message or a long meeting recording and it returns the text. Without a subscription the bot gives 60 minutes per month for free.

Built-in dictation and Potok

Based on Microsoft and Apple help pages.
ToolWhere it worksRussian languageWhere speech is recognized
Windows voice typing (Win+H)Windows 10 and 11Windows 11 onlyMicrosoft cloud, does not work offline
macOS DictationmacOSyeson the device or on Apple servers
PotokWindows, macOS, Linuxprimary languagecloud, servers in Russia, does not work offline
Mark Vesyoly — founder of Potok. Writes about Russian voice input: how to dictate in work apps, what gets in the way and how to fix it.
© Potok · Vladimir Olegovich Lipatnikov · INN 502915859701 · Public offer Comparisons · FAQ · About Free voice input · Voice input for AI chats · Data and Russian privacy law · Voice input on Mac · Bot plans · 14 days of Pro for a review