A bot that transcribes voice messages and audio in MAX

1. Open the bot in MAX

   Use the button on this page or MAX search. The bot tells you right away what it accepts, and the first two transcripts need no sign-up on the site.

2. Forward a voice message or file

   A voice message from any chat is easiest to forward: long-press it, choose “Forward” and pick the bot. Attach audio, video and documents as a regular file, and paste a meeting link as plain text.

3. Get the text and key points

   The transcript arrives in the same chat as a file, a PDF by default: split by speaker with a timestamp on every line. Buttons under it pull out key points and tasks, answer a question about the recording or save the text in another format.

There are two other routes. A short message is quicker to read with MAX’s built-in “T” button — see above. And if a recording is too large or you want to edit it straight away — naming speakers, fixing words while listening, building a report from a template — save the file from MAX and upload it on the site.

How to link the MAX bot to your site account

The bot works without linking too, but then MAX recordings live only in the chat. Once linked, they appear in the web workspace next to files uploaded on the site, and bot settings and your plan are shared with the site. You link once: from then on everything you forward to the bot shows up on the site by itself.

1. On the site, open Settings → Integrations and click “Connect Telegram or MAX”.
2. The site shows a one-time link valid for 10 minutes. Click “Open the bot in MAX” — the chat with the bot opens.
3. Start the bot in MAX. The status on the site updates by itself, no need to go back and refresh.
4. If you had already opened the chat with the bot before, send it /start followed by the code the site shows.

MAX has its own transcription: a “T” button sits next to a voice message and shows its text under the recording. It's a quick way to read a short message instead of listening to it. The Vibe2Text bot doesn't replace it — it does a different job: it turns a long conversation into a document you can work with. Think of them as two tools: the button answers “what did they say to me”, the bot answers “what did we agree and what happens next”.

The “T” button is enough when

• the voice message is short and recorded by one person
• you only need to read it and reply
• there is nothing to save or forward

Use the Vibe2Text bot when

• several people talk and it matters who said what
• you need the outcome: key points, tasks, agreements and deadlines
• it's an audio file, a meeting video or a meeting link, not a voice message
• the text has to be saved as DOCX or PDF and found later on the site

Where it works

in the MAX messenger on phone, desktop and web

What it accepts

voice messages, audio, video, documents with a recording and meeting links

File size

up to 2 GB per file — our MAX bot’s own limit, and it tells you when a file is over it; up to 4 GB on the site

What it returns

text by speaker with timestamps; on request — key points, tasks and answers

Export formats

DOCX, PDF and TXT right in the chat

Account

shared with the site and the Telegram bot once linked

Who said what

Speech is split by speaker — “Speaker 1”, “Speaker 2” — with a timestamp on every line. Put real names in the web workspace and they change across the whole text at once.

Outcomes and tasks

A half-hour stand-up becomes a few lines of substance and a task list: what to do, who owns it and by when, if the call said so.

Answers from the recording

Ask the bot what price was agreed or what the contractor promised — it answers from the transcript and shows the moment it was said.

A document, not a message

Save the finished text as DOCX, PDF or TXT right in the MAX chat, ready to forward or attach to a task. The same recording is in your web workspace.

Example: what comes back for a voice message

A made-up example for a roughly three-minute voice message where two people discuss a delivery. Names and details are invented; the structure of the reply matches the real bot.

Key points

• The delivery moves to Tuesday: the warehouse can't pack the order by Friday.
• The client has to be told today and confirm the new date.
• If the client objects, a partial shipment on Friday is on the table.

Tasks

• Tell the client about the delay — today before lunch
• Ask the warehouse what can ship partially — by Thursday
• Update the delivery date in the order once the client replies

The summary and tasks are switched on with a button under the transcript; by default the bot sends only the text.

A manager on the move

A day in work chats brings a dozen three-to-five-minute voice messages. Listening in a taxi without headphones is awkward, and missing an instruction is costly. Forward the batch to the bot: by evening every one has its text and a short list of what's being asked of you, so you can reply to the point without replaying anything.

A sales manager

A client dictated order requirements in a voice message and then added two more. The bot turns them into text you can paste into the deal card, and the key-points button pulls out dates, amounts and terms. If two people spoke, lines are labelled — you can see what you offered and what the client asked for.

Students and teachers

A lecture or consultation recorded on a phone goes to the bot as a file. Back comes text with timestamps, easy to search, and notes with the main ideas. A question like “which topics are on the exam” can go straight to the bot: it finds the answer in the transcript.

Journalists and researchers

An hour-long interview recorded on a phone becomes text by speaker that only needs proofreading. A quote is easy to check against the audio by timestamp, and the DOCX goes straight to the editor. Long recordings are better sent as a file than as a voice message — the original sound quality is kept.

A parent

In the parents' chat the teacher recorded a four-minute voice message: the school trip, what to bring, the deadline for the money. The bot turns it into text, and the tasks button leaves a short list that's easy to forward to the other parent who has no time to listen.

A specialist on site

A site manager or engineer dictates snag notes in voice messages on the go, hands busy. In the evening the bot turns them into text and a to-do list with owners, if names were said — all that's left is attaching the document to the report or task.

What people send the bot

• Voice messages from colleagues and clients in MAX work chats
• Stand-ups and calls recorded on a phone
• Videos of meetings, webinars and open lessons
• Meeting links from Yandex Telemost, MTS Link and SaluteJazz
• Lectures, interviews and consultations an hour long or more

What comes out

• Manager's voice message → task list
• Stand-up recording → minutes by speaker
• Lecture audio → notes with timestamps
• Interview → text for an article
• Telemost meeting link → transcript and outcomes

MAX is the messenger more and more Russian work chats, parents’ chats and client conversations have moved to, so that’s where the voice messages you have to listen to arrive too. The bot lives where the recordings are: no saving the file to your phone, digging through downloads and uploading it to a website — a voice message is forwarded with one tap and a meeting video is attached like any file. If you prefer Telegram, the same bot is there too; both share one Vibe2Text account, and recordings from either messenger land in one history on the site. Many people already have MAX installed for work or a school chat, so transcription doesn’t mean installing one more app or signing up for an unfamiliar service.

How the MAX bot differs from the Telegram bot

It’s the same service with one account: transcription, speaker labels, key points, questions about the recording and export work the same way. The differences are small. Bots in MAX can’t react to messages, so you won’t see an emoji on yours: the bot reports receipt and completion with separate messages, and you can close the chat meanwhile. Our MAX bot’s per-file size limit is lower than in Telegram — the numbers are in the table above. Pick the messenger where your recordings already are: forwarding voice messages within one app is easier than moving files between apps.

You don't have to wait for someone to send a call recording. Send the bot in MAX a Yandex Telemost, MTS Link, SaluteJazz or Kontur Talk meeting link — it joins the call as a separate participant, records the conversation and, when it ends, sends a transcript by speaker to the same chat. Along the way the bot updates a single status message: joining, recording, done. Participants see the bot in the list, so tell them about the recording in advance — it's polite and settles personal-data questions. Key points and tasks for such a meeting come from the same button as for voice messages, and the recording shows up in the web workspace if your account is linked.

The bot only gets what you send or forward to it: it can’t see your other chats. We store recordings and their text in Russia, and they are visible only to you — and, once your account is linked, in your web workspace. Delete a transcript you no longer need there: the source and the text leave the account. If a recording contains other people’s phone numbers, addresses or other personal data, get their consent before sending it — that rule doesn’t depend on the messenger or on who transcribes. Your recordings aren’t shown to other users; a transcript is shared only explicitly — through a link you create yourself in the web workspace.

Can the bot transcribe a voice message from a MAX group chat?

Yes, if you forward it to the bot: long-press the message in the group, choose “Forward” and pick the bot. There's no need to add the bot to the group itself — members won't see that you transcribed anything.

Can the bot see my chats in MAX?

No. The bot only gets the messages you send or forward to it yourself: private and group chats it isn’t in are closed to it. There’s no need to add the bot to a group to transcribe something — just forward it the voice message, and the other members won’t know.

Does the MAX bot work on desktop and on the web?

MAX isn't phone-only: there's a desktop app and a web version in the browser. The bot works the same everywhere — forward a voice message from a laptop, or attach a meeting video saved on disk like any file. Handy when a meeting recording sits on a work computer: no need to move it to your phone. The bot's reply reaches all your devices at once, and with a linked account the same transcript opens in the web workspace for work on a big screen.

What’s better to send: a voice message, audio or video?

Voice messages are easiest to forward as they are — MAX keeps them in a quality that suits recognition. A meeting recording is better sent as the original audio or video file rather than retold by voice: that keeps every participant and the pauses between lines the bot uses to separate speakers. Video is heavier than audio, so for a long meeting that’s too large, sending just the audio track is enough. And if the meeting hasn’t started yet, send the bot its link — it records and transcribes it itself.

How do I get an accurate transcript of a voice message?

Accuracy depends on sound far more than on length. Speak into the phone's microphone rather than to the side, and avoid speakerphone where you can: echo and background music confuse both recognition and speaker labels. Record a meeting with one device in the middle of the table rather than forwarding fragments from several phones. If the conversation matters, send the original file: forwarding through several chats doesn't hurt the sound, but re-recording it on a second phone does. After transcription, check names, company names and figures — that's where people and neural networks alike make mistakes.

Can I put names on the speakers?

Yes, in the web workspace: open the transcript and name the participants — “Speaker 1” and “Speaker 2” change across the whole text at once. Your MAX account needs to be linked to the site for that, and then the bot’s recordings show up in the workspace by themselves. The bot tells voices apart on its own, but it doesn’t know or guess names: only you provide them.

What if the bot got some words wrong?

Open the transcript in the web workspace and correct the text there: the timestamp on every line helps you find and replay the right fragment quickly. Mistakes are most common in surnames, company names, rare terms and figures. If a recording matters and the sound is poor, send the original file rather than one re-recorded from a speaker — that’s the most effective thing you can do before transcription.

Which languages does the bot understand?

The bot recognises Russian and English best, including accented speech and mixed phrases where Russian sentences carry English terms. Other languages are supported too, with lower accuracy, so proofread those recordings more carefully. The language is detected automatically, there's nothing to choose before sending. If a recording is full of industry jargon, product names or surnames, check them in the finished text: rare words are where recognition slips most.

What can I do with a finished transcript?

The most common path is to press the key-points button and forward a short summary to colleagues instead of a three-minute voice message. If you need a document, save it as DOCX or PDF right from the chat. For a long meeting, open the recording in the web workspace: there you can put names on speakers, fix individual words while listening to that exact fragment and build a report from a template. And to find one detail in an hour-long call, just ask the bot — it answers from the text and tells you when it was said.

How does a team use the bot?

The bot links to a personal account: each person connects their MAX to their own profile on the site, and their transcripts go to their history. For colleagues to see each other's recordings, work in a shared workspace in the web app — that's also where access is set up. Voice messages from different chats then collect in one place instead of getting lost in private chats, and any meeting's text and outcomes can be found even if you weren't there.

Before you send a recording

• The voice message is forwarded or the file is attached
• A long conversation is sent as the original file, not re-recorded from a speaker
• If others' personal data is on the recording, they agreed to it
• Names, titles and terms are checked in the finished text
• Your account is linked in Settings → Integrations so recordings show up on the site

If something goes wrong

The bot says the file is too large

Our MAX bot has a per-file size limit and tells you itself when a file is over it. Compress the video, send only the audio track or upload the recording on the site, where the limit is higher.

The account link didn't work

The one-time link lasts 10 minutes. Go back to Settings → Integrations, get a new one and open it right away. If the chat with the bot is already open, send it /start with the code from the site.

The transcript is almost empty

That happens when a recording is music, silence or speech far from the microphone. Check that the voice is clearly audible and send the original file instead of a re-recording from a speaker.

The transcript is taking a long time

Long recordings take longer, and the bot only writes when it's done. If there's no reply for noticeably longer than usual, message the bot again or check the recording in the web workspace if your account is linked.