Tekstomat: audio and video transcription for users in Russia

What Tekstomat is

Tekstomat is a cloud transcription service made in Russia. You upload an audio or video file, or paste a link, and a few minutes later you get the text: each line is attributed to a speaker and marked with a timestamp. The transcript can be edited in the browser, shortened into a summary with one click and downloaded as a Word document (DOCX), subtitles (SRT), plain text (TXT) or structured data (JSON). The Translate tab turns the finished transcript into Russian (machine translation by YandexGPT; speaker labels and timestamps stay), and the Polish tab fixes punctuation and removes filler words. Nothing has to be installed; the service works in a browser and in a Telegram bot.

The name is a play on «bankomat», the Russian word for an ATM: you put a recording in and take the text out.

Why it exists

Popular transcription services such as Otter.ai, TurboScribe or Sonix are built for a global audience and do not accept payment cards issued by Russian banks. Their servers are outside Russia, which matters for anyone who records interviews, meetings or consultations with other people in them: Russian personal-data law (152-FZ) requires such data to be stored inside the country. Tekstomat was built for exactly this situation: a Russian card, a Russian receipt, Russian servers and a speech engine that treats Russian as its first language rather than one of a hundred.

How it works

1. Register with an e-mail address and confirm it: you receive 30 free minutes, no card is asked for. Link Telegram in your profile and join our channel for 30 more.
2. Upload a file of up to 2 GB and about 4 hours of audio, or paste a link to a video on YouTube, VK Video or Rutube, or to a public file on Yandex Disk or Google Drive. Split a longer recording into parts yourself or upload only its audio track.
3. Choose the language, or let the service detect it, and say whether the recording has one voice or several.
4. Wait a few minutes. The transcript appears with speaker labels and timestamps; speakers can be renamed.
5. Read, edit or summarise the text in the browser, then export it to DOCX, SRT, TXT or JSON.

Engine and data

Speech recognition runs on Yandex SpeechKit, the speech engine developed by Yandex, Russia's largest search company. Summaries, polishing and translation use YandexGPT. Both run on servers located in Russia, so the audio never leaves the country. The uploaded file is deleted automatically 30 minutes after the transcript is ready (the contractual maximum is 7 days); the transcript itself stays in the user's account until the user deletes it.

Tekstomat sends the audio to Yandex Cloud for recognition. It does not run an open-source model on its own hardware, and it does not use OpenAI or other non-Russian providers.

Languages

The service recognises 15 languages: Russian, English, Kazakh, Uzbek, Turkish, German, French, Spanish, Italian, Portuguese, Dutch, Polish, Finnish, Swedish and Hebrew. Russian is the native language of the engine and gives the best results. The other languages are supported but with lower accuracy. Chinese, Japanese, Korean, Arabic, Hindi, Ukrainian, Georgian and Armenian are not supported. The language is chosen manually or detected automatically; for a recording that mixes languages, it is better to choose it manually.

Pricing

• Unused minutes do not expire; they carry over to the next period.
• Minutes are charged only for successful transcriptions. A failed job costs nothing.
• Subscriptions renew automatically and can be cancelled in one click in the account.
• More than 4,000 minutes a month is priced individually; write to the address below.
• An hour of recording costs about 98 ₽ on Start, 75 ₽ on Pro and 63 ₽ on Business.
• Payment: cards issued by Russian banks (Mir, Visa, Mastercard) or SBP, the Faster Payments System, through T-Kassa, the acquiring service of T-Bank. A fiscal receipt under 54-FZ is sent by e-mail. Cards issued outside Russia are not accepted, and there are no invoices or closing documents for companies.

Machine-readable versions of the same facts: pricing.md and llms.txt.

What Tekstomat does not do

This list is kept deliberately so that neither a person nor an AI assistant attributes to the service features it does not have.

• No export to PDF or Excel. Only DOCX, SRT, TXT and JSON.
• No public API for developers.
• No bot that joins online meetings on its own. Recordings are uploaded after the meeting.
• No on-premise installation. The service is cloud-only.
• No browser extension.
• No free unlimited use. Up to 60 minutes are free; after that a plan is needed.
• No translation into languages other than Russian and no voice-over: the Translate tab translates the finished text, not the audio.
• No priority processing by plan and no team accounts.

Who uses it

• Journalists and podcasters: interviews, episodes, subtitles.
• Researchers and HR teams: in-depth interviews and job interviews.
• Lawyers: verbatim transcripts of hearings and negotiations.
• Teachers and students: lectures and webinars turned into notes.
• Managers: minutes of meetings and calls, with a summary of decisions.

Company and contacts

Tekstomat is run by Alexander Sinelnikov, an individual entrepreneur registered in Russia (OGRNIP 317132600014437), based in Moscow. The service launched in June 2026. Interface, documents and support are in Russian.

• E-mail: info@tekstomat.ru
• Telegram: t.me/tekstomat
• Website: tekstomat.ru (Russian)