Make SRT and VTT subtitles from AMR files

By the Potok teamUpdated ~2 min read

AMR files become SRT and VTT subtitles when you send them to Potok in Telegram, and 2 files come back as downloads. AMR is a compact voice format, so quiet or distant speakers may leave gaps in the captions. Read the SRT once and fix any gap before you publish it.

Turn a recording into text

Drop an audio or video file, or a voice message, into the chat with @potok_now_bot; the recognized text is sent back there.

Go to @potok_now_bot in Telegram

Export to TXT, MD, DOCX, PDF, SRT and VTT · files up to 2 GB · about 30 languages · speakers and timestamps on request

In short

  • AMR storage files begin with the magic string #!AMR followed by a newline; AMR-WB uses #!AMR-WB.
  • Android lists AMR-NB in 3GPP (.3gp) and AMR (.amr) files at 4.75 to 12.2 kbps sampled at 8 kHz.
  • A single-channel AMR file holds only a magic number before the speech frames (audio only), so an SRT stays a separate file.
  • The bot reads .amr files of up to 2 GB directly; Potok adds no conversion step.

AMR at a glance

PropertyValue
KindSpeech codec for mobile phones
Media typesaudio/AMR, audio/AMR-WB
What it holdsAdaptive Multi-Rate speech: AMR-NB (.amr) and AMR-WB (.awb)

How to make subtitles from AMR files

  1. Hand the file to the bot. Find @potok_now_bot in Telegram and add your .amr file as a message. The bot takes up to 2 GB; very long recordings are better uploaded with the mini app's Upload button.
  2. Keep the format in mind. A single-channel AMR file holds only a magic number before the speech frames (audio only), so an SRT stays a separate file.
  3. Get the subtitle file. Press the SRT or VTT export button; each caption carries a start and an end time. Then attach the file to the video and check how it lines up.
  4. A useful detail. Android lists AMR-NB in 3GPP (.3gp) and AMR (.amr) files at 4.75 to 12.2 kbps sampled at 8 kHz.

What the subtitle file looks like

Example of SRT captions

[00:00:01,000 --> 00:00:03,400] Caption 1: Hello, it's Peter from the dentist's office.

[00:00:03,600 --> 00:00:06,100] Caption 2: Your appointment is moved to Thursday morning.

[00:00:06,300 --> 00:00:09,000] Caption 3: Please call back if that time does not work.

[00:00:09,200 --> 00:00:11,800] Caption 4: Hello, I am calling about the parcel that never arrived.

Captions written by us to show the layout. They are not output from a real AMR file.

Potok at a glance
TopicAnswer
InputVoice messages, audio and video
Short digestA summary button under the transcript gives a short digest
Output formatsTXT, MD, DOCX and PDF for text; SRT and VTT for subtitles
TranslatingThe text keeps the language of the audio and is not translated
Spoken languagesAbout 30 languages with good published results; a few more with warnings

Typical AMR slips

Where things go wrong
SlipResultRemedy
Hoping the AMR file carries its own captionsIt only has speech frames, so a player shows noneKeep the SRT as a separate file
Sending a quiet AMR call with distant speakersLow volume speech gives gaps and wrong words in captionsRecord closer to the speaker or raise the level
Using the SRT times after converting the AMR elsewhereA trimmed conversion shifts the caption timingRequest the SRT from the exact file you will play
Sending an AMR file recorded in a moving carEngine noise creates wrong words in the captionsRe-record somewhere quiet or review each caption after receiving

Next steps

Terms used here are explained in the glossary. Handy helpers live in tools; apps are on the download page.

Questions and answers

Can I send AMR files straight to Potok?

Yes. .amr files up to 2 GB go to the bot as they are; there is no conversion step before recognition.

Will the captions be burned into my AMR file?

No. You get a separate .srt or .vtt text file, and your original AMR file stays unchanged.

Which apps and devices create AMR files?

Android MediaRecorder offers an AMR_NB output format (Android MediaRecorder AMR_NB).

Sources

Each source was opened by hand on the date shown. Pages change, so open the source if a detail matters to you.

  1. RFC 4867: RTP payload format and file storage format for AMR and AMR-WB, IETF · checked
  2. Supported media formats, Android Developers · checked

Run your own recording through it

Drop an audio or video file, or a voice message, into the chat with @potok_now_bot; the recognized text is sent back there.

Go to @potok_now_bot in Telegram