Frequently
asked questions

All common audio and video formats: MP3, WAV, OGG, FLAC, M4A, MP4, MOV, WebM, and more. The bot also handles voice messages, video notes, and audio files uploaded from your device. File size depends on your plan: up to 300 MB on FREE, up to 2 GB on START, PRO, and POWER. File length is capped at 5 minutes on FREE, 30 minutes on START, and is unlimited on PRO and POWER.

Yes. Forward a voice message or video note to @taktextbot — the bot extracts the audio track and returns a transcript with timestamps. Learn more on the Telegram voice-to-text page.

Yes. Tap "Forward" on any voice message or video note in any chat and choose TAK! TEXT. The bot never accesses the original chat — it only processes the forwarded audio. Whether the original sender's info is visible depends on Telegram's forwarding settings.

YouTube, Vimeo, Dailymotion, SoundCloud, Dropbox, Twitch, Rumble, Bandcamp, Mixcloud, and other supported platforms — over 20 platforms in total. Send a link to the bot and it will download the audio and transcribe it. Link transcription is available on PRO and POWER. Whether a specific video or audio can be transcribed depends on the platform's current restrictions and the recording's own privacy settings — private or restricted content may not be accessible.

More than 90 languages. For most languages, both modes — "Speed" and "Quality" — are available. For rare languages, or languages that are often misdetected automatically, you can pick the language manually in the bot's settings — this noticeably improves accuracy.

The bot automatically detects up to 48 speakers and labels each line as "Speaker 1:", "Speaker 2:", and so on. It's useful for interviews, meetings, and group recordings. Diarization works on every plan, including FREE.

Not yet — for now the bot works only in private chats. Group chat support is in development and will be available soon.

"Speed" mode processes audio fast — short voice messages are ready in seconds. "Quality" mode is more accurate, supports more languages, and handles noisy audio, accents, and overlapping speech better, though it can take a little longer. Both modes are available on every plan and switch with one tap in the bot's settings.

On clean speech — above 95% in "Quality" mode and 92–95% in "Speed" mode. Accuracy drops with heavy background noise, low-quality recordings, and overlapping speech from multiple people. Language also matters — common languages score higher; for rare ones you can manually set the recording's language in settings.

In the bot, open the settings menu (the "⚙️ Settings" button or the /settings command). Choose "Recognition mode" and toggle between "Speed" and "Quality". The setting persists for all future transcriptions.

Open the bot menu → "Settings" → "Transcription settings" → "Recognition language". Pick a language from the list or leave "Auto-detect". The chosen language behaves differently depending on the mode:

In "Quality" mode it's a hint to the bot: it boosts accuracy for the selected language, but audio in other languages is still recognized. Highest accuracy = "Quality" + an explicitly chosen language.

In "Speed" mode it's a strict constraint: results come back as fast as possible, but audio in other languages won't be recognized.

After transcription, three tools are available. Summary — a brief 3–5 bullet-point recap (free on every plan). Translation — into 13 languages. Ask the text (Q&A) — ask a question about the transcript and get an answer grounded in the text.

Yes. The brief recap (summary) is free on every plan, including FREE. Translation and Q&A use AI requests: 10 free per month on FREE, 50 on START, unlimited on PRO and POWER.

13 languages: English, Spanish, French, German, Portuguese, Italian, Arabic, Turkish, Persian, Russian, Ukrainian, Uzbek, and Kazakh. Translation is triggered by the "Translate" button after a transcript is ready. We're expanding the list — if a specific language is missing, reach out to support.

Ask a question about the transcript and the bot will answer based on the text. For example: "What did we agree on?", "When is the next meeting?", "Which numbers were mentioned?". It's handy for long recordings when you need to find specific information quickly. Questions must be about the transcript content — the bot won't answer off-topic questions.

Four plans. FREE — 15 minutes of transcription per month (30 minutes in your first month), files up to 5 minutes, 10 AI requests. START (€3.19/mo) — 300 minutes, files up to 30 minutes, 50 AI requests. PRO (€9.69/mo) — 1000 minutes, no per-file length limit, unlimited AI. POWER (€29.49/mo) — 3000 minutes, unlimited AI. Full details on the pricing page.

Two options for international users. Telegram Stars — the primary method, works in a single tap right inside the Telegram app, available internationally. Bank cards (Visa, Mastercard, UnionPay, and others), Apple Pay, Google Pay, and PayPal — via Stripe. Availability of specific methods depends on your region. Local payment rails are also offered in some markets directly inside Telegram.

Yes. How you cancel depends on the payment method.

Bank card (Stripe): in the bot menu → "💳 Manage subscription" → open the generated link to your Stripe customer portal and cancel auto-renewal.

Telegram Stars: in the bot menu → "💳 Manage subscription". You can also manage it from Telegram's Stars or subscription settings.

In both cases, access remains active until the end of the paid period. Unused minutes do not roll over to the next month.

Packages are extra minutes that stack on top of your subscription. If you've used up the 300 minutes on START, you can top up with a package and keep working without upgrading to PRO. Packages stay until you use them — no monthly charges.

You can also buy packages on FREE — in that case the START plan limits apply (files up to 30 minutes, up to 2 GB).

It depends on the plan. On FREE — up to 5 minutes per file. On START — up to 30 minutes. On PRO and POWER there is no length cap within the 2 GB maximum file size.

Very long files — especially in "Quality" mode — can take a while to process. The result arrives in the bot with a notification, so you can close the bot in the meantime.

Audio and video files are deleted immediately after processing — the bot doesn't store them. Transcripts are kept for up to 24 hours so the AI tools can work, then they're deleted automatically and permanently.

As a European company, we're required to comply with EU court orders. But since we don't store audio at all, and transcripts live no longer than 24 hours, in practice there's effectively nothing to hand over — even with an official request.

So that AI tools can run on the transcript. Summary, Q&A, and translation all run on the stored transcript. If it's already gone from the server, these features are technically impossible. The same goes for the "Download transcript" button — for long recordings that don't fit in a single Telegram message, downloading won't be available after 24 hours.

The transcript message itself stays in your chat with the bot — only you can see it, and we no longer have that data on our servers.

The core TAK! TEXT infrastructure runs on Hetzner Online GmbH, with data centers in Germany (EU). Certain processing operations may be performed by connected providers, including outside the EU, under appropriate safeguards (SCCs, DPA). Full list — in our privacy policy. More on security — on the dedicated page.

Yes. TAK! TEXT processes data in accordance with GDPR. The data controller is sershiko (Netherlands, KVK 42031706). You can request deletion of your data or a copy of it — email privacy@taktext.com. Details — in our privacy policy.

Transcripts are deleted automatically after 24 hours. Audio files are not stored — they're deleted immediately after processing.

If you need data deleted sooner or right away — the fastest path is to message the support bot. Official GDPR deletion requests are handled through privacy@taktext.com.