How to Turn a Telegram Voice Message into Text
Telegram transcribes voice messages itself, for Premium users and a few a week for everyone else. When that's enough, and how to keep a timestamped transcript of a long recording.
Short answer
Tap the transcribe button on the voice message: Telegram Premium includes it, and other users get a small weekly allowance. For a long recording, a file you want to keep, or more than the allowance, forward the voice message to @SubGrabBot after connecting your account at subgrab.com/telegram-bot. It returns a timestamped transcript at 1 credit per started hour.
Voice messages are quick to send and slow to receive. A three-minute voice note can't be skimmed, searched or quoted, and a twenty-minute one recorded in a meeting is worse. Turning it into text fixes all of that, and Telegram has two ways to do it: its own, and sending it somewhere that keeps a proper transcript.
Telegram's own voice-to-text
Telegram can transcribe a voice message where it sits: tap the button on the voice message and the text appears under it. According to Telegram's API documentation:
- With Telegram Premium, it's part of the subscription, with no weekly allowance to run out of.
- Without Premium, you get a set number of transcriptions a week, and each message can be up to a set length. When Telegram opened the feature to free users in 2023, the allowance was two a week.
- In a boosted group, members without Premium can transcribe there without using their own allowance.
For a short voice note you only need to read once, that's the right tool, and it costs nothing extra.
When you need more than that
Telegram's transcription shows text under a message. Some jobs need more:
- A long recording. A lecture, a meeting or an interview sent as a voice message or an audio file.
- A file to keep. Text you can download as Word, PDF, TXT, Markdown, SRT, VTT, CSV or JSON, with a timestamp on every line.
- A summary. The key points and a timeline of a long recording, not just the words.
- More than the weekly allowance, if you don't have Premium.
Send it to the SubGrab bot
SubGrab's Telegram bot, @SubGrabBot, transcribes whatever you forward to it on your SubGrab account, the same way it transcribes an uploaded file.
1. Open subgrab.com/telegram-bot while signed in and tap Open in Telegram, then press Start in the chat that opens. That connects the chat to your account; the link works once, for 30 minutes.
2. Forward the voice message to @SubGrabBot. It also takes round video messages, audio files and video files, up to 2 GB.
3. The bot sends you the transcript's link when it's ready, and the transcript stays in your dashboard.
The spoken language is detected from the audio itself, so a voice message in another language, or one that switches language halfway, needs no setting.
What it costs, honestly
1 credit per started hour of audio, up to 10 hours per file, and every new account starts with 2 free credits. That makes a 40-minute recorded meeting 1 credit, which is good value. It also makes a 20-second voice note 1 credit, which isn't, so for short notes Telegram's own button is the better choice.
If a transcription fails, its credits come back automatically.
What the bot keeps
Only the ID of your chat with the bot, so it knows which account a file belongs to: not your name or phone number. A file you send is downloaded, transcribed and deleted when the job ends; the transcript is titled with its caption or file name, as an upload keeps its file name. Send the bot /disconnect at any time to unlink the chat. The details are in our privacy policy.
Frequently asked
Is Telegram's voice-to-text free?
Partly. Telegram's documentation says users without Premium can transcribe a set number of voice messages a week, each up to a set length; it was two a week when the feature reached free users in 2023. Premium users have no weekly allowance to run out of, and members of a boosted group can transcribe there without using theirs.
Can the SubGrab bot transcribe audio files, not just voice messages?
Yes: voice messages, round video messages, audio files such as MP3, M4A or OGG, and video files, forwarded or sent to the bot directly, up to 2 GB and 10 hours of audio.
Which languages work?
SubGrab transcribes 90+ languages and detects the spoken one from the audio itself, so a voice message in Spanish, Hindi or Russian needs no setting. Accuracy is about 95% on clear English speech and varies more for languages with less training data.