SaluteSpeech Bot: Voice to Text

Someone sent you a three-minute voice message and you are on the subway, in a meeting, or simply do not feel like listening. SaluteSpeech Bot turns such a message into text. Send it, skim it in half a minute, reply.

The bot was made by Sber on top of its own speech recognition platform. More than half a million people use it. 🎙

Two ways to transcribe a voice message

First: forward or send a voice message straight to the bot in a private chat. Second: add the bot to a group, and it will transcribe the members' messages.

The /group command explains how to add the bot to a chat. /help opens answers to the main questions.

What is known about the features

• Price. According to media reports at launch, transcription in the bot is free.
• Punctuation. The bot adds punctuation, so the text reads like a normal message and not a wall of words.
• Format. Voice messages and audio files are accepted. The launch news mentioned a size of up to 20 MB, but check current terms in /help.
• Languages. The official help mentions Russian and English.

Where it really helps

Voice messages are easy for the one speaking and hard for the one listening. Text can be skimmed, and you can find a number or an address and copy it.

It is especially handy with poor sound, in a noisy place, or when you need the gist of a long message fast. People with hearing loss can also use such a bot to follow conversations. 👂

Transcription in a group chat

If your work group loves voice messages, the bot spares everyone else from listening. The text appears for all under the message.

Add it deliberately, though: transcription goes through Sber's service, and personal data of other people can end up in a group. Warn the members first.

Getting a more accurate result

1. Speak clearly. Do not rush and do not swallow word endings.
2. Avoid noise. Wind, music and traffic make the result worse.
3. Hold the phone closer. The microphone should catch your voice and not the background.
4. Recheck names and numbers. Rare surnames, brand names and figures are recognized worst.

About privacy

Any transcription means your audio is processed on someone else's servers. Do not send the bot voice messages with passwords, bank details or other people's secrets.

Read the processing terms in Sber's help before sharing anything sensitive. 🔐