Telegram voice to text bot without code

A Telegram voice to text bot in nana800 takes three blocks: the trigger catches a voice message, the Speech to text block transcribes it with OpenAI, and the bot replies with the text. You need a bot token from @BotFather and your own OpenAI key, with no code. Telegram blocks cost $1 per 1,000 runs, transcription costs twice as much, and OpenAI bills the recognition itself to your OpenAI account.

Build it in the builder

What you can automate

  • The bot answers a voice message with its text, handy when you cannot listen to the recording
  • Voice requests from customers become text and go to your team group in Telegram
  • Dictate a note or a task and get it back as text in a private chat with your bot
  • Transcribe audio files, videos and video notes up to 20 MB: the block reads mp3, m4a, wav, ogg, flac and mp4 video
  • Get product names and people right by listing them in the model hint, and set ru,en for speech in two languages
  • Pass the transcript further down the scheme: to a spreadsheet, a CRM or an AI model for a short summary

How to set it up

  1. 1

    Create a bot and an OpenAI key

    Send /newbot to @BotFather and copy the bot token. Create an OpenAI key in the API keys section of OpenAI Platform. OpenAI charges your own account for transcription, so it needs a positive balance.

  2. 2

    Add the incoming message trigger

    Create a scheme and add the Telegram message from user to bot trigger. Paste the token and choose message as the event type. nana800 copies every voice message the bot receives to its own storage and puts the file link into the audio_urls list of the trigger.

  3. 3

    Add the transcription block

    Add the Speech to text block and paste your OpenAI key. In the audio or video link field, insert audio_urls from the trigger and add [0] at the end of the path to take the first link. Set the language to ru or en and keep the gpt-transcribe model.

  4. 4

    Skip messages without voice

    If people also send text, put a condition block before transcription with one rule: the audio_urls link exists (the EXISTS operator). Connect transcription to the check mark exit, so plain text messages never reach transcription and do not cause errors.

  5. 5

    Send the text back

    Add the Telegram message send block with the same token. Put chat_id from the trigger into the chat ID field, the text field of the transcription block into the message, and pick the plain_text format. To reply right under the voice message, put message_id from the trigger into the reply field.

  6. 6

    Turn the scheme on and test it

    Turn the scheme on and send the bot a short voice message. If no answer comes, the builder marks the failed block and the run log shows the reason: a wrong key, an empty link or a file over 25 MB. While the scheme is on, it costs $0.0042 per hour.

Pricing

A run of a Telegram block costs $0.001, a run of a Speech to text block costs $0.002. Every month the balance is topped up to $3.50 for free, an enabled scheme costs $0.0042 per hour.

Services in this connection

  • Telegram

    Build a Telegram bot without code: reply to messages, add buttons, send leads and alerts to a chat, group or channel. Free balance every month, no card needed.

  • Speech to text

    Turn audio, video and Telegram voice messages into text with OpenAI, no code: calls, meetings and interviews. Free balance every month, no card needed.

Frequently asked questions

In nana800 a voice message runs the trigger and the reply at the basic block price, $1 per 1,000 runs. Transcription costs twice as much. An enabled scheme costs $0.0042 per hour. Your balance is topped up to $3.50 for free every month, but a bot kept on around the clock spends most of that on scheme hours. OpenAI bills the recognition itself to your account at its own prices.

No. The Speech to text block works only with your own key: nana800 downloads the recording from the link and sends it to OpenAI for transcription. You create the key in OpenAI Platform, and the account needs a positive balance.

Voice messages and audio files that Telegram hands to bots, up to 20 MB, with links in audio_urls. Videos and video notes work too, their links are in video_urls. The transcription block takes mp3, m4a, wav, webm, ogg, flac and mp4 up to 25 MB and detects the format itself. Telegram gives no link for files over 20 MB.

Set the language of the recording, for example ru or en: this makes recognition more accurate and faster. List product names, people and terms in the model hint field. For a recording in two languages enter ru,en, which only the gpt-transcribe model understands.

Yes. Choose with_timestamps in the output field and the whisper-1 model: you get the text and a list of segments with start and end times in seconds. Other models do not return timestamps. OpenAI will shut down every model except gpt-transcribe on February 26, 2027.

Build your first scheme for free

Sign-up is free and needs no card. Every month the balance is topped up to $3.50 for free, enough for about 3,500 runs of basic blocks.

Build it in the builder