Trim AMR audio files online. Cut voice memos and phone recordings to the exact segment you need.
Process files in seconds with our optimized servers
Set exact start and end points with frame accuracy
Maintain original quality with smart re-encoding
AMR (Adaptive Multi-Rate) is a 3GPP-standard speech codec built for the GSM and UMTS phone networks — the audio that carried 2G and 3G calls, MMS voice messages, and the default Android voice-recorder format on early Android (2.x and 3.x) before Google switched to AAC-in-3GP and later AAC-in-M4A. AMR-NB targets 8 kHz mono speech at 4.75-12.2 kbps; AMR-WB (also called G.722.2, files often .awb) targets 16 kHz mono speech at 6.6-23.85 kbps. The format is tiny — a one-minute AMR-NB clip is roughly 60-90 KB — and tuned for human voice, not music. Common reasons to trim:
.3gp container. Pull the speech segment out for transcription or to forward as a standalone .amr without the surrounding container.For a different output after trimming, see AMR to MP3, AMR to WAV, or AMR to OGG.
| Property | AMR-NB | AMR-WB | MP3 |
|---|---|---|---|
| Sample rate | 8 kHz | 16 kHz | 8-48 kHz |
| Bitrate range | 4.75-12.2 kbps | 6.60-23.85 kbps | 32-320 kbps |
| Channels | Mono only | Mono only | Mono or Stereo |
| Tuned for | Narrowband telephone speech | Wideband speech (HD voice) | Music + speech |
| File ext | .amr | .awb (or .amr) | .mp3 |
| 1-minute size | ~60-90 KB | ~80-180 KB | ~240-2400 KB |
| Universal playback | Limited (Android, VLC, ffmpeg) | Limited (VLC, ffmpeg) | Yes (every device) |
| Best for | 2G call recordings, legacy archives | HD-voice / VoLTE captures | Sharing and music |
AMR's strength is byte-for-byte the smallest meaningful speech file you can ship; its weakness is that iOS, macOS Finder Quick Look, Windows Media Player, and most browsers don't play it natively. Convert to MP3 (or WAV) only when the recipient can't open .amr.
| Bitrate mode | Use case | Notes |
|---|---|---|
| AMR-NB 4.75 kbps | Lowest-bandwidth voicemail, archival of 2G calls | Audibly compressed but intelligible |
| AMR-NB 7.40 kbps | Default for many Android voice recorders | Good clarity / size tradeoff for narrowband speech |
| AMR-NB 12.2 kbps | Highest AMR-NB quality (also called GSM-EFR) | Best AMR-NB quality before switching to AMR-WB |
| AMR-WB 12.65 kbps | Common HD-voice / VoLTE call mode | Wideband speech, noticeably more natural than NB |
| AMR-WB 23.85 kbps | Highest AMR-WB quality | Top of the AMR ladder; switch to Opus or AAC for further gains |
| MP3 64 kbps mono | Re-encode for universal playback | Smallest MP3 still clean for speech |
| MP3 128 kbps mono | Re-encode for shareability + headroom | Drop to mono since AMR is mono-only |
If the recording must stay native, keep AMR. If anyone on iOS, Windows, or a browser needs to play it, re-encode to MP3 mono in step 3.
Trimming alone is a precise byte-range cut, but AMR is a frame-based speech codec, so XConvert decodes and re-encodes the kept segment with the same AMR-NB or AMR-WB mode it came in at. The output stays at the source bitrate by default — quality only changes if you opt into a different bitrate or codec in step 3. Pick the highest bitrate the source uses (12.2 kbps AMR-NB, 23.85 kbps AMR-WB) and the loss versus the original frames is negligible for speech.
AMR-NB is the original narrowband codec — 8 kHz sample rate, 4.75-12.2 kbps, the audio of 2G GSM calls and the early Android voice memo format. AMR-WB (also published as G.722.2) is wideband — 16 kHz, 6.60-23.85 kbps, the codec behind 3G/VoLTE "HD Voice." Files use .amr for narrowband and .awb for wideband, but Android often writes both as .amr. XConvert detects the mode automatically; if you re-encode in step 3 you can keep it the same or switch.
Native playback is patchy. iOS doesn't open .amr from Files or Mail without a third-party app, macOS Finder Quick Look refuses it, and Windows Media Player skips it. VLC plays AMR on every platform, and ffmpeg / mpv handle it from the command line. For frictionless playback elsewhere, re-encode to MP3 or WAV in step 3, or use AMR to MP3 afterward.
AMR frames are 20 ms each (160 samples at 8 kHz for NB, 320 samples at 16 kHz for WB). XConvert decodes and re-encodes around your start and duration values, so the output is sample-accurate within the codec — typically within a single 20 ms frame of the timestamp you entered. For voicemail, court excerpts, and call-recording edits that's tighter than human ear can detect.
AMR-NB and AMR-WB are mono-only codecs by spec; the format has no stereo mode. The mic on the phone that captured it was almost certainly mono too. If you need a stereo output for a NLE timeline, switch the codec in step 3 to MP3, AAC, FLAC, Opus, or WAV — those let you choose Mono or Stereo (the second channel will be a duplicate of the mono source unless you mix in something else).
Yes. Add multiple trim ranges — each pair of start time + duration produces a separate output file. Useful for pulling three relevant exchanges out of a 45-minute call recording, splitting an all-day push-to-talk session log into individual transmissions, or cutting a multi-message voicemail dump into one file per caller.
There's no fixed cap. Trimming runs in your browser, so the practical limit is your device's available memory. AMR is so compact that even a 12-hour call-recorder log is usually under 60 MB at AMR-NB 7.4 kbps, and a full-day field-recorder roll fits comfortably. Multi-hour AMR files trim in seconds once uploaded.
Trim first. AMR's whole point is small files, so trimming first keeps the working set tiny and means the optional MP3 re-encode in step 3 (or via AMR to MP3 afterward) only has to process the seconds you kept, not the whole hour.
Most car head units and Bluetooth speakers don't decode AMR — they want MP3, AAC, or WAV. Trim first to keep the slice you want, then in step 3 switch the codec to MP3 (128 kbps mono is plenty for speech) or use AMR to WAV for the most universal playback.