When you send a voice note to @DoublegramScribeBot on Telegram, you rarely stare at a blank screen until the end. The bot uses streaming transcription: the text updates progressively while the audio is still being processed.
This is automatic in @DoublegramScribeBot — no toggle in Settings, no extra command.
What you see
- You send a voice message.
- Scribe immediately replies with 🎙️ (processing indicator).
- Within a second or two, partial text appears.
- The message keeps updating as more of your speech is recognized.
- When finished, the full transcription is shown — prefixed with 🎙️.
Longer voice notes benefit most: you start reading early instead of waiting for the entire recording to finish.
Private chat vs groups
Streaming works in both places, but Telegram displays updates differently:
| Chat type | How updates appear |
|---|---|
| Private chat | Live draft messages — text refreshes in place as Scribe receives chunks from the transcription engine |
| Group | The bot's reply message is edited in place — same message, growing text |
In private chat, the final draft already contains the complete text — Scribe skips a redundant final edit when drafts were sent successfully.
Forum topics
Streaming works inside forum topics in your private chat with Scribe. Updates appear in the same topic thread where you sent the voice note.
Rate limiting (why updates aren't instant)
Scribe throttles Telegram updates slightly so it does not hit API limits:
- Updates are batched — not every micro-chunk triggers a visible refresh
- Small pauses between updates keep the experience smooth on Telegram's side
You still see progressive text; updates may arrive in short bursts rather than letter-by-letter.
When streaming is not available
If streaming fails mid-process, Scribe shows Unable to transcribe this message on the processing reply. Resend the voice note — a fresh attempt usually works.
In rare deployments where streaming is disabled server-side, Scribe falls back to non-streaming mode: you see 🎙️ until the full transcription is ready, then one final update with the complete text. Most users always get streaming.
Streaming vs AI chat
Streaming applies to transcription only — converting speech to text. When Scribe generates an AI reply (chat, reminders, history search), that uses a separate response flow. Voice transcription streaming does not cost AI credits; AI responses may use credits depending on your plan.
Try it: Send a 20–30 second voice note describing your day. Watch the text build before you finish listening to yourself.
Quick reference
- Default: streaming on
- Private: live drafts
- Groups: message edits
- Indicator: 🎙️ prefix throughout
- No user setting to enable/disable
Related: understand group transcription limits in Group Transcription Limits and Premium on Telegram.