TAK! TEXT: turn speech into text

A coworker sends a seven-minute voice note while you are stuck in a meeting. Or you have a lecture recording and need three key points from it. TAK! TEXT handles this inside Telegram: forward the recording and read the text shortly after. 🎙

It understands more than 90 languages, detects the language on its own and needs no sign-up. Around 11,000 people use it each month.

What you can send it

• Voice notes and video messages from any chat, just forward them.
• Audio and video files in MP3, WAV, OGG, FLAC, M4A, MP4, MOV and WebM.
• Links to YouTube, TikTok and Instagram videos, plus files from cloud storage. This needs a paid plan or a minute package.
• Group chats. Add the bot and mention it when someone needs a transcript.

Text is only the first layer 📝

The bot labels up to 48 different speakers and adds timestamps roughly every 30 seconds, so finding a moment in a long meeting is quick.

AI tools sit on top. Summaries are free on every plan. Translation and questions about the recording ("what did we decide on the budget?") use AI requests from your allowance.

There are two modes, Speed and Quality. Quality is more accurate and covers more languages. For noisy audio or strong accents, set the language by hand in /settings. Export the result as TXT or PDF.

The price list

Prices are in dollars and euros. For subscriptions starting on or after August 15, 2026:

• FREE, $0. 30 minutes in your first 30 days, then 15 minutes every 30 days. Recordings up to 5 minutes, files up to 300 MB, 10 AI requests.
• START, $3.49. 300 minutes a month, files up to 2 GB, 50 AI requests, links.
• PRO, $8.29. 1,000 minutes and unlimited AI requests.
• POWER, $23.99. 5,000 minutes for teams and heavy users.

Prefer no subscription? One-time packages range from 70 to 2,000 minutes, last 365 days and never auto-renew. On the free plan an active package opens up links and 2 GB files while the minutes last. Check /premium for current options.

What happens to your recordings 🔒

A fair question, especially for work calls. Audio and video are deleted right after processing. Transcripts are kept encrypted for up to 24 hours, then erased for good.

Servers are in Germany and the service states it follows GDPR. The company behind it is registered in the Netherlands and based in Amsterdam.

Tips for a cleaner transcript

1. Record close to the mic. Distance and echo cost more accuracy than any mode setting.
2. Pick Quality for interviews, lectures and anything important.
3. Set the language manually when there is an accent or lots of jargon.
4. Double-check names and numbers. Clean speech gets above 95% accuracy, but surnames and amounts deserve a second look.