Potok or SpeakFlow: what to choose for voice input in Russian

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

Potok (potok.now) and SpeakFlow are Russian voice input services with payment in rubles. Potok inserts text with a hotkey into any field on Windows, macOS and Linux, while SpeakFlow works on Windows and in the browser. Potok's free plan is 2500 words per week, SpeakFlow's is 30 minutes per month; the paid plan costs 690 ₽/mo at both.

How SpeakFlow and Potok are alike

They have quite a lot in common. Both services are made in Russia for Russian-speaking users: behind SpeakFlow, according to the offer on speakflow.ru, is sole proprietor Evgeny Sergeevich Shintyapin (self-employed), and behind Potok is a sole proprietor in Russia too. The idea is the same: to replace part of manual typing with voice where saying is faster than typing. The plans are arranged similarly: a free entry and a paid level that removes the limits.

The price matches too. The SpeakFlow “Personal” plan and Potok Pro cost the same, 690 ₽ a month, and both services take payment in rubles, with no foreign currency cards and no intermediaries. This sets the pair apart from foreign tools, whose prices are in dollars. There is no independent measurement of the accuracy of one service against the other, so we do not claim “who is more accurate” here.

Where they are really different: a time limit against a word limit

The first difference is in the product itself. On its website SpeakFlow describes more scenarios around speech: transcription of meetings, turning audio and video into text, dictation in Russian with insertion of a translation into English, correction of punctuation and spelling. It works on Windows and in the browser. Potok is narrower and does one thing: you hold Ctrl+Space, speak, release, and the ready text lands where the cursor was, in any app on Windows or macOS. Potok has no upload of ready files.

The second difference concerns the unit of the free limit. For SpeakFlow it is 30 minutes of dictation per month, for Potok 2500 words per week per device. One metric counts recording time, the other the volume of the resulting text, and you cannot convert one into the other without assumptions about speech pace and pauses. So it would be dishonest to declare either limit more generous. Estimate your own mode. Do you dictate often and a little at a time through the week? The metric in words is closer. Rare long sessions once every few days fit more easily into a monthly budget of minutes.

There is a practical detail about money. The paid SpeakFlow plan removes the minute limit for the same 690 ₽/mo. Potok Pro is unlimited too, and when paid for a year the price drops to 499 ₽/mo. If you count the yearly cost of the subscription, this is a noticeable difference.

Tuning for Russian speech and data

Potok runs on its own model trained for Russian speech: names, professional terms, switching between Russian and English inside one phrase. By our measurement on the GolosTest benchmark, accuracy is a WER of 2.1% (the share of wrongly recognized words; the lower, the better). This is our figure from our own test, and there is no direct comparison with SpeakFlow in Russian. The response time is about 0.4 seconds, and the dictation speed is about 900 characters per minute. There is a personal dictionary for terms and brands and voice snippets.

Processing runs on servers in Russia. For companies there is a 152-FZ mode and on-prem deployment on their own servers, when the sound does not leave the perimeter. For private correspondence this is a small thing, for a bank, a clinic or a legal department it is often the deciding point.

There are fewer verifiable data on SpeakFlow. Russian is stated there as the main language, and recognition is in the cloud too. In which country the servers are and on which engine recognition works is not disclosed on the website, and there are no terms for corporate data (152-FZ, on-prem) there either. We will not fill this in for SpeakFlow: if the jurisdiction of the data matters to you, check its privacy policy.

Potok is a cloud service, recognition runs on servers in Russia; it does not work without the internet.

What to choose for your task

If you need to insert text quickly all day (messages in messengers, notes, emails, documents, prompts in an IDE) and want it to work in any app without switching windows, Potok's flow fits directly. It is also more practical if you work on macOS or in a company that needs 152-FZ and servers in Russia. You can test the scenario on yourself in 14 days of full access with no card.

If the main task is to go through a past meeting, turn accumulated recordings into text or get a translation of the dictated text into English right away, this is exactly what SpeakFlow states. The monthly limit in minutes then fits rare bulky sessions more logically. One more argument for it: work in the browser without installation.

The price of the paid plans is the same, so the scenario decides. It is reasonable to try the free level of both: Potok has 2500 words per week, SpeakFlow 30 minutes per month. In a couple of weeks it will be clear which flow and which limit metric is more convenient on your texts.

Frequently asked questions

Built-in dictation and Potok

Primary sources: support.microsoft.com · support.apple.com