Initializing... drag & drop files here
Supports: MKV
This tool throws away the video in a Matroska (.mkv) file and keeps only the soundtrack, re-encoding it as AMR — the speech codec built for mobile telephony. Be clear-eyed before you start: MKV normally carries full-bandwidth audio (AAC, AC-3, DTS, Opus or FLAC), and AMR is the narrowest target on the site. Speech survives it and shrinks enormously; music does not. This walk-through covers the one step that trips everybody up on this page — you have to choose the AMR variant before the bitrate list has anything in it — plus what happens to an MKV's extra audio and subtitle tracks along the way.
.mkv onto the page or click "+ Add Files" to pick it from your computer. Several files can be queued and converted with the same settings..amr. Files upload over an encrypted connection, are processed on our servers, and are deleted automatically a few hours later — no sign-up, no watermark.The Constant Bitrate dropdown on this page is populated by whichever AMR variant is active, and nothing is selected when the page loads. That is not a stuck control — it is waiting on step 2. Choose an Audio Codec first and the list fills in immediately:
| Property | AMR Narrow Band (AMR-NB) | AMR Wide Band (AMR-WB) |
|---|---|---|
| Sample rate on this page | 8000 HZ (locked) | 16000 HZ (locked) |
| Speech bandwidth | roughly 300–3,400 Hz | roughly 50–7,000 Hz |
| Modes offered here | 8, from 4.75 to 12.2 KBPS | 9, from 6.60 to 23.85 KBPS |
| Also standardised as | 3GPP AMR (ETSI) | ITU-T G.722.2, approved July 2003 |
| Marketed as | ordinary mobile call quality | HD Voice |
| Channels | Mono only | Mono only |
| File signature | begins #!AMR |
begins #!AMR-WB |
| Pick it when | a handset, MMS or IVR system expects classic AMR | the receiver documents wideband support |
Matroska is deliberately a multi-track envelope — the format's own documentation describes it as holding "many audio, video and subtitles streams" with selectable tracks and chapter entries, and it is now standardised as RFC 9559 (October 2024). A .amr file is the opposite: one mono speech stream and nothing else. Everything below is dropped, not hidden:
| What is in the MKV | What ends up in the .amr |
|---|---|
| Video track | Discarded entirely |
| Default audio track | Decoded and re-encoded as AMR |
| Additional audio tracks (commentary, dub languages) | Not included — one track only |
| Soft subtitles | Not included; AMR has no subtitle concept |
| Chapters, tags, cover art, attached fonts | Not included |
| Stereo or 5.1 imaging | Downmixed to a single mono channel |
.amr" — Desktop players are patchy with AMR, and some that read narrowband files reject wideband ones because the header differs. VLC opens both; otherwise convert onward with AMR to MP3.AMR is the wrong destination whenever fidelity matters — songs, film soundtracks, podcasts, anything musical. No mode setting rescues that; the bandwidth is gone by design. Encrypted or partially downloaded MKV files may also fail to decode at all, and an MKV whose audio stream is itself broken produces a silent result no matter what you pick. The honest default for most people extracting sound from a Matroska file is MKV to MP3 for listening or MKV to WAV for editing; reach for AMR only when a feature phone, voicemail platform, IVR menu, or telephony system specifically demands a .amr file.
Because the AMR bitrate modes are defined per variant and no variant is preselected. AMR Narrow Band and AMR Wide Band have completely different mode tables — eight rates versus nine, at different sample rates — so the dropdown has nothing valid to offer until you choose one under Audio Codec. Pick a variant and the modes appear at once. If you would rather not think about modes, leave File Compression on Quality Preset; that path derives a bitrate from the codec's own range for you.
Narrowband (AMR-NB) samples at 8 kHz and reproduces roughly the 300–3,400 Hz telephone band across eight modes from 4.75 to 12.2 KBPS. Wideband (AMR-WB) samples at 16 kHz, covers roughly 50–7,000 Hz across nine modes from 6.60 to 23.85 KBPS, and is the codec behind "HD Voice" — it is also published by the ITU as Recommendation G.722.2, "Wideband coding of speech at around 16 kbit/s using Adaptive Multi-Rate Wideband (AMR-WB)". Wideband sounds clearly better on voice; narrowband is what the oldest handsets and telephony equipment expect.
By its first bytes — AMR files declare their variant in a plain ASCII magic string at the start of the file. We checked encoder output both ways: a narrowband file begins #!AMR and a wideband file begins #!AMR-WB, even though both use the .amr extension. That is also why a player that only understands classic narrowband AMR can silently refuse a wideband file you made on this page; if a device rejects your output, re-run it as Narrow Band before assuming the conversion failed.
No, and no setting changes that. AMR is a speech codec: it models a human vocal tract rather than storing a general waveform, and it discards everything outside the voice band before it even reaches the bitrate you chose. Instruments lose their harmonics, bass vanishes, and cymbals turn to hiss. Spoken content — interviews, narration, lecture recordings, dialogue — stays perfectly intelligible, which is exactly what AMR was designed for. For music extracted from an MKV, use MKV to MP3 or MKV to FLAC.
The one the container marks as the default audio track. This page has no track selector, so a film with an English track plus commentary and dub languages produces a single .amr from whichever stream is flagged default, and the rest are dropped. If that is not the track you need, remux the MKV first with your preferred stream set as default. The same applies to soft subtitles and chapters: AMR has no way to store them, so they simply do not carry over.
Small enough that duration, not quality, drives the number. At the 12.2 KBPS narrowband maximum, a minute of audio is about 90 KB; at 4.75 KBPS it is closer to 35 KB; at the 23.85 KBPS wideband maximum roughly 175 KB. Those figures follow directly from the bitrate, because AMR modes are constant-rate. An hour of narrowband speech therefore lands around 5 MB — which is exactly why voicemail systems and messaging platforms standardised on it.
Yes. Open Trim in Advanced Options — it starts on Unchanged — and enter a Start time and a Duration, either in seconds or as HH:MM:SS.sss. Only that span is decoded and encoded, which is much faster than converting a two-hour file and cutting it afterwards. If you would rather cut the video first and keep a video copy of the excerpt, use the Video Cutter and convert the result.
None. AMR is an audio-only file format with no video capability whatsoever, so the picture is decoded only far enough to reach the audio stream and is then discarded. What you get back is a single mono speech file. If you actually wanted a smaller video rather than an audio extract, convert to MKV to MP4 or compress the original with Compress MKV instead.
Your MKV is uploaded over an encrypted connection, converted on our servers, and both the upload and the resulting .amr are deleted automatically a few hours afterwards. Nothing is shared or made public, there is no sign-up, and no watermark or audio tag is added to the output. The original file on your own device is never modified — the AMR is written as a new file.