How to get text from MP3

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

The Potok Telegram bot @potok_now_bot gets text from MP3: send a file in private messages, and it transcribes the speech into text; a short recording comes back at once, a long one as a file. A large file is easier to upload in the bot's mini app, up to 2 GB by dragging. Recognition is free for 60 minutes per month, and there is no translation into other languages.

Did you record a conversation on a voice recorder, download a podcast or receive a meeting recording in MP3, and you need text? The easiest way is to send the file to the Potok Telegram bot @potok_now_bot: a separate transcription program is not needed, the bot does everything.

How to get text from MP3 step by step

1. Open the Telegram bot @potok_now_bot and press Start (the /start command) if you write to it for the first time.
2. Send the MP3 file in private messages as an attachment, as an ordinary document or as audio.
3. For a large file the bot's mini app is more convenient: drag in a file of up to 2 GB, and if the connection breaks, the upload continues from where it stopped.
4. A short recording comes back as text right in the chat (or on the screen in the web version), a longer one as a separate file.
5. The format of the automatic reply (TXT or DOCX) is chosen beforehand in the bot settings (/settings).
6. Timestamps and speaker labels, Speaker 1 and Speaker 2, can also be turned on in the bot settings (/settings).
7. With the buttons under the finished text you get a short summary, a list of tasks by speaker and an answer to your own question about the recording at once, and the Other format button exports the same text to Markdown, PDF and SRT and VTT subtitles.

How to check and correct the transcript

Before you use the text, it is worth checking it against the recording. The bot proofreads long transcripts itself: it places punctuation and capital letters and removes obvious recognition errors, but it does not shorten the text or retell it. The bot's mini app has a player: a click on any phrase in the text plays the recording from exactly that place, so it is easier to check a doubtful word by ear. All transcripts are saved in the My recordings section, and you can open and check them at any time without uploading the file again. If several people speak in the recording, it is worth turning on speaker labeling, Speaker 1 and Speaker 2, in advance: replacing one line with another in the finished text would have to be done by hand.

When the bot will not fit

The Potok bot will not help if the recording is not in Russian: it recognizes only one language, and there is no translation. The bot will not tell you the real names of the participants either: in the labeled text they appear as Speaker 1 and Speaker 2. Transcription works only on an already finished recording: the bot cannot join a live call and turn speech into text on the fly. It does not process links to Instagram and TikTok either, and of the video platforms it understands only YouTube, VK Video and RuTube. A broader overview of transcription services is in the comparison of transcription services. It is a cloud service: without a network recognition will not start, and all processing runs on Potok's servers in Russia.

Frequently asked questions

Primary sources: speech2text.ru · audio-2-text.ru · turboscribe.ai · turboscribe.ai