Initializing... drag & drop files here
Supports: MPG, MPEG
MPEG-1 and MPEG-2 program streams (.mpg, .mpeg) were the dominant video formats from the mid-1990s through the early 2000s — Video CDs, DVDs, MiniDV camcorder exports, and TV capture cards all used them. Their audio tracks (typically MPEG-1 Layer II or AC-3) are CD-quality stereo, which is overkill for any application that just needs the spoken content. AMR (Adaptive Multi-Rate) is the speech codec that GSM, UMTS, and VoLTE phone networks use — it compresses voice down to 4.75-23.85 kbps by modeling the human vocal tract instead of sampling the waveform. Common reasons to extract MPEG audio as AMR:
If you need the audio in a more universal format instead, see MPEG to MP3 or MPEG to WAV. For a different speech-extraction workflow, AMR to MP3 goes the other direction.
| Property | MPEG (source) | AMR (output) |
|---|---|---|
| Type | Video container with audio track | Audio-only, speech codec |
| Audio quality | CD-quality stereo (MP2 or AC-3) | Speech-only mono |
| Sample rate | 44.1 kHz or 48 kHz | 8 kHz (NB) or 16 kHz (WB) only |
| Frequency range | Full audible spectrum (up to ~22 kHz) | 300 Hz - 3.4 kHz (NB) / 50 Hz - 7 kHz (WB) |
| Typical bitrate | 128-384 kbps audio | 4.75-12.2 kbps (NB), 6.60-23.85 kbps (WB) |
| 1-minute file size (audio portion) | ~1-3 MB | ~35-90 KB (NB), ~50-180 KB (WB) |
| Best for | Video playback, music, archival | Voice ringtones, MMS, telephony, IVR |
| Universal device playback | Wide (DVD players, VLC, browsers) | Narrow — phones, VLC, Audacity, specialist apps |
| Mode | Bitrate | Best for |
|---|---|---|
| AMR-NB | 4.75 kbps | Smallest possible — emergency-channel voice, MMS over EDGE |
| AMR-NB | 7.40 kbps | Typical GSM voicemail quality |
| AMR-NB | 12.2 kbps | Default — best AMR-NB quality, used by GSM full-rate calls |
| AMR-WB | 6.60 kbps | Smallest wideband — usable HD voice for tight bandwidth |
| AMR-WB | 12.65 kbps | Balanced wideband for clear dialogue at low size |
| AMR-WB | 23.85 kbps | Maximum AMR-WB quality — closest to natural voice in AMR |
Pick AMR-NB if the target system is a feature phone, MMS gateway, GSM voicemail, or any pre-2010 mobile network — those endpoints almost always assume narrowband. Pick AMR-WB (HD Voice) if the destination is a modern VoLTE-capable phone, a SIP softphone, or an Asterisk/FreeSWITCH PBX configured for wideband. AMR-WB at 12.65 kbps sounds noticeably clearer for dialogue than AMR-NB at 12.2 kbps because it captures up to 7 kHz of audio versus 3.4 kHz, but it's not universally accepted on older handsets.
No. AMR is a speech-only codec built around the ACELP algorithm — it models the vocal tract and discards everything outside the speech band. Music, sound effects, and ambient noise come out muffled, warbly, and chorus-like, and percussion turns to mush. If your MPEG has any musical content worth keeping, convert to MP3 or M4A instead. Use AMR only when the source is pure speech.
Two reasons stack. First, you're discarding the video track entirely — that's typically 80-95% of the MPEG file. Second, you're re-encoding the remaining audio from a 128-384 kbps full-spectrum stereo stream down to a 4.75-23.85 kbps speech-band mono stream, which is another 10-50× reduction. A 30 MB MPEG with 60 seconds of dialogue ends up as a 35-90 KB AMR-NB file or a 50-180 KB AMR-WB file.
Always mono. AMR has no stereo mode — the encoder downmixes the source automatically. This is fine for ringtones, MMS, IVR, and telephony, which are all mono pipelines anyway. If you need stereo speech output, convert to MP3 or AAC instead.
8 kHz for AMR-NB and 16 kHz for AMR-WB. There's no other option — these are the rates the codec specification defines. The source MPEG audio (44.1 or 48 kHz) is automatically downsampled before encoding. If you need a different sample rate, AMR is not the right format for your use case.
Yes. Drop in multiple .mpg or .mpeg files at once and each converts in parallel within your browser session. Output downloads as individual AMR files or as a single ZIP. Useful when batch-prepping IVR prompts, building a corpus of speech samples, or converting a folder of archived dialogue clips for a phone-system migration.
Yes. Use the trim section to enter a start time and duration. Both accept seconds (12.5) or HH:MM:SS.sss format (00:01:30.500). This is useful for cutting a single line of dialogue out of a long MPEG, removing the intro music before speech begins, or chopping a multi-minute clip down to the few seconds you actually want as a ringtone.
About 35 KB at AMR-NB 4.75 kbps. That mode is intelligible but noticeably lower fidelity — fine for transmission over a 2G EDGE link, an emergency-channel voice memo, or an MMS attachment that has to fit under a tight carrier limit. For anything you want to actually listen to comfortably, AMR-NB 12.2 kbps (90 KB/min) or AMR-WB 12.65 kbps (95 KB/min) is the realistic floor.
Modern iOS and Android no longer ship a native AMR player in the default Files / Music app — Apple removed AMR support around iOS 11, and stock Android dropped it in recent releases. The file still works in VLC, Audacity, and most third-party media players, and it remains valid for MMS, ringtone slots on older phones, and any backend system (Asterisk, FreeSWITCH, IVR platforms) that explicitly expects AMR. If you need a file that just plays on a current consumer phone, use MPEG to MP3 instead.