A voicemail becomes text the same way any recording does: save it as a file, upload it, read the transcript. The only thing that makes voicemail its own problem is the audio β a phone line delivers mono sound band-limited to roughly 300β3400 Hz, and callers say the important part, the phone number, faster than anything else in the message. Everything below is about that.
Rather than claim a quality figure, this guide lets you check one. Play a real voicemail, read the unedited transcript our pipeline produced from it, and see how the same engine behaved when the audio was degraded to genuine telephone conditions.
Voicemail to text at a glance
| What you need | The voicemail saved as an audio file |
| Getting the file | iPhone share sheet, Android export, Google Voice download, or carrier voicemail-to-email |
| Formats accepted | 27 audio and video extensions β including the .amr and .3gp many tools reject |
| Conversion needed first | None. Phone-quality audio uploads as-is |
| Free use | First 5 minutes with no account β longer than almost any voicemail |
| Typical processing time | Seconds. A 16-second message took 4.65 s in the test below |
| Exports | TXT, Markdown and JSON free; DOCX, PDF, SRT, VTT on paid plans |
| Confidential messages | Transcribe on-device with the Mac app β audio never uploads |
- Hear a real voicemail and its transcript
- How to get the voicemail off your phone
- Which voicemail formats work (including AMR)
- Does phone audio hurt accuracy? We measured it
- Carrier and iPhone voicemail transcription
- Voicemail to text software compared
- How to translate a voicemail to text
- Do you need a voicemail to text app?
- Where voicemail transcription goes wrong
- Test methodology and results
- FAQ
- Sources
Hear a real voicemail and its transcript
Press play. The transcript below the player is the unedited output of our production pipeline on this exact file β nothing was corrected by hand. Click any line to jump to that moment in the audio.
Two things worth noticing. The engine punctuated a caller who never paused for punctuation, including the self-interrupting *"I've called, I want to say, five times today"*. And it kept the sentence boundary at *"picked up once."* even though the speaker ran straight into the next thought β which is what makes a voicemail transcript skimmable rather than a wall of words.
How to get the voicemail off your phone
This is the actual friction. Transcription takes seconds; extracting the file is where people get stuck. Four routes, easiest first:
Carrier voicemail-to-email. If your carrier or business phone system can forward voicemails to your inbox, turn it on. The message arrives as a WAV or MP3 attachment already saved as a file β no phone gymnastics at all. This is the single best option if it is available to you, and it is the one most people do not know they have.
iPhone. Open the Phone app, go to the Voicemail tab, tap the message, then tap the share button and save it to Files (or send it to yourself). The file lands as `.m4a`. If the share button is missing, your carrier's visual voicemail is not delivering the audio to the device β use the voicemail-to-email route instead.
Android. The path depends on your voicemail app rather than on Android itself. In Google Voice, open the message on the web and use the download option; you get an `.mp3`. In a carrier visual-voicemail app, look for save, export or share on the individual message. Files from Android recorders often arrive as `.amr` or `.3gp` β see the format section, because that is exactly where other tools fail.
Last resort: re-record. If nothing else works, play the voicemail on speaker and record it with a second device. Quality drops, and it is the only method here that adds a real accuracy penalty β the microphone picks up the room as well as the message.

Which voicemail formats work, including AMR
Voicemail arrives in a narrower set of formats than general audio, and one of them is a common failure point.
| Format | Where voicemail in this format comes from | Accepted |
|---|---|---|
| .m4a | iPhone Phone app share sheet | Yes |
| .mp3 | Google Voice download, carrier email attachment | Yes |
| .wav | Carrier voicemail-to-email, office phone systems | Yes |
| .amr, .3gp | Android voicemail and call-recorder apps | Yes β no conversion needed |
| .aac, .ogg, .opus | Messaging apps and voice-note forwards | Yes |
| .aiff, .caf | macOS recordings of a played-back message | Yes |
| .mp4, .mov | Screen recording of a voicemail playing | Yes β audio track extracted |
AMR is the one that matters. It is the codec Android voice and call recorders have used for years, and several browser transcription tools reject it, telling users to convert to MP3 first. That advice costs a conversion step and a second generation of lossy compression on audio that is already band-limited. We accept `.amr`, `.amr-wb`, `.3gp` and `.3g2` directly β the whitelist our upload validation enforces covers 27 extensions in total, so the practical answer for voicemail is that whatever your phone produced will upload.
Does phone audio hurt accuracy? We measured it
This is the question underneath every voicemail-to-text search: *the recording is terrible, will it even work?* Rather than quote a percentage, we ran the test.
We took a clean 44.1 kHz stereo recording of a spoken message and produced a second copy encoded to genuine telephone conditions β mono, 8 kHz, band-passed to 300β3400 Hz, through the mu-law codec carriers actually use. Both went through the same engine, same settings.
| Studio original | Telephone-band copy | |
|---|---|---|
| Audio | 44.1 kHz stereo | 8 kHz mono, mu-law, 300β3400 Hz |
| Transcript text | Identical, word for word | |
| Segments returned | 7 | 3 |
| Processing time | 4.65 s for 16.16 s of audio | |
Band-limiting the audio to a phone line changed nothing about the words. What it changed was segmentation: the engine merged short phrases into longer blocks, because the cues it uses to place a boundary β brief pauses, breath β are exactly what the band-pass filter attenuates. Practically, that means a voicemail transcript may arrive as fewer, longer lines than the same words spoken into a good microphone, but not as different words.
This is one clip, not a benchmark. It does establish the thing worth knowing: telephone bandwidth is not the reason a voicemail transcript comes back wrong. The reasons are further down, and they all live on the caller's end.
Carrier and iPhone voicemail transcription
Before converting anything by hand, check what you already have β and understand why it often disappoints.
Carriers generally sell voicemail transcription as a paid add-on to visual voicemail rather than including it. Availability, price and quality vary by carrier, plan and country, so check your carrier's own page rather than a comparison article. The common complaint is not that it costs money but that the transcript is a single unpunctuated run of text with the phone number mangled β carrier transcription has historically run on older, cheaper speech models than the ones a dedicated transcriber uses.
iPhone has two separate things that are easy to confuse. Live Voicemail transcribes an incoming message on-device, in real time, as the caller is speaking, so you can read it and decide whether to pick up; its language and region coverage is narrow, so check whether it appears in your Settings at all. Separately, whether a message you already have shows a transcript in the Voicemail tab depends on your carrier's visual voicemail, not on Apple.
Google Voice transcribes automatically and lets you download the audio, which makes it the friendliest starting point: you get a free first-pass transcript and the original `.mp3` to re-transcribe properly if the first pass mangled a name.
Use a dedicated transcriber when the carrier transcript is not good enough, when you need the message as a document, or when the message is in a language your carrier does not transcribe.
Voicemail to text software compared
Published figures from each vendor, checked July 2026. They change β confirm before relying on them:
| Free tier | AMR / 3GP | Playable proof | Exports | |
|---|---|---|---|---|
| TranscribeNext | First 5 min, no account; 3 files/day at 50 MB with a free account | Yes, direct | Yes β on this page | TXT, MD, JSON free; DOCX, PDF, SRT, VTT paid |
| AudioScribe | Up to 5 min, no signup; 25-min files on an account | States AMR is not supported | No | TXT, SRT, VTT on a paid account |
| Ticnote | Max 50 MB; signup for larger files | Not published | No | Not published on the tool page |
| Carrier add-on | Usually paid, bundled with visual voicemail | N/A β transcribes in the network | No | None β text appears in the app |
| Google Voice | Free with a Google Voice number | N/A β transcribes in the service | No | Audio download as .mp3 |
Two honest notes about our own row. The 5-minute free ceiling is generous for voicemail specifically β carriers cap messages well below it β but it is a preview limit, not an unlimited free tier. And Word and PDF export are paid; free accounts get TXT, Markdown and JSON, which is enough to read, search and paste a voicemail but not to produce a formatted document.
How to translate a voicemail to text
Do it in two steps, not one, and the result is markedly better.
1. Transcribe in the spoken language first. Set the language manually instead of leaving auto-detect on β a short, band-limited message gives detection very little to work with, and a misdetected language produces confident nonsense rather than an obvious error.
2. Correct names and numbers against the audio while you still have it open. These are the tokens a translator cannot repair, and a wrong digit survives translation looking perfectly plausible.
3. Then translate the text. A clean transcript is clean input; a translation engine working directly from phone audio compounds two error sources and hides which one failed.
The exception is a caller who switches language mid-message, which is common in voicemail and which neither auto-detection nor a single-language setting handles well. Transcribe with the dominant language set, then fix the switched passage by hand β it is usually one sentence.
Do you need a voicemail to text app?
For converting messages you already have: no. A browser does the job on any device, and installing something to transcribe one file is effort you do not need to spend.
An app earns its place in two cases. The first is volume β if you want every incoming voicemail transcribed automatically rather than exporting and uploading each one, that is what an app or a carrier add-on is for. The second is confidentiality: a client's voicemail, a medical result, an HR message. In that case the argument is not convenience but that the audio never leaves the machine β TranscribeNext for Mac transcribes on-device on Apple Silicon, so a sensitive message is never uploaded anywhere.
Where voicemail transcription goes wrong
As the test above showed, phone bandwidth is not the culprit. These are, roughly in order of how often they bite:
The pattern: everything on this list happened before the audio reached you. Nothing you choose downstream fixes a caller who mumbled a number in a moving car β which is exactly why you should check names and digits rather than trusting any transcript, ours included.
Test methodology and results
Test date: 27 July 2026
Source clip: βWoman Leaving a Phone Messageβ by Sample_Me, published on Freesound under CC0 1.0 (public-domain dedication). 16.1 s, originally 44.1 kHz stereo.
Telephone-band copy: produced with ffmpeg β downmixed to mono, resampled to 8 kHz, band-passed 300β3400 Hz, encoded through the mu-law codec, then packaged for the web. This is the file the player above serves, so the transcript you read is the transcript of the audio you hear.
Engine: TranscribeNext production pipeline, Whisper large, language set to English, diarization off, default profile β identical settings for both versions.
Result: transcripts word-for-word identical. Studio original returned 7 segments; telephone-band copy returned 3. Processing took 4.65 s for 16.16 s of audio.
Transcript shown: unedited engine output. Line start/end times are measured; the word-level highlight is interpolated inside each measured line, because our current backend returns segment-level timings rather than per-word ones.
This is one clip on one day, not a benchmark. It is published so the claim can be checked rather than believed.
Last tested: 27 July 2026
FAQ
How do I convert a voicemail to text? Save the voicemail as an audio file, then upload it to a transcription tool and read the transcript. On iPhone, open the Phone app, go to Voicemail, select the message and use the share button to save it as an .m4a file. On Android, export the message from your voicemail or call-recorder app, or download it from Google Voice as an MP3. Many carriers can also email voicemails to you as an audio attachment, which is the easiest route of all because the file arrives ready to upload.
Can I transcribe a voicemail for free? Yes, for short messages. TranscribeNext transcribes the first 5 minutes of any file free without an account, which covers essentially every voicemail β the format is designed to be short. A free account raises it to 3 files a day at 50 MB each and exports TXT, Markdown and JSON. PDF and Word exports are on the paid plans.
Does poor phone audio hurt voicemail transcription accuracy? Less than people expect. We took a clean 44.1 kHz recording of a spoken message, re-encoded it to true telephone conditions β mono, 8 kHz, 300-3400 Hz band-pass, through the mu-law codec carriers use β and ran both versions through the same engine. The transcripts came back word for word identical. Only the segmentation changed: the studio file split into 7 segments, the telephone file into 3. What does hurt accuracy is background noise on the caller's end, crosstalk, and mumbled names and digits.
Can voicemail-to-text handle AMR files from Android? Ours does. AMR and AMR-WB are the codecs Android voice recorders and call recorders have used for years, and several online transcription tools reject them outright β which is why people are told to convert to MP3 first. TranscribeNext accepts .amr, .3gp and .3g2 directly, along with 24 other audio and video extensions, so no conversion step is needed.
Can I translate a voicemail to text in another language? Yes, in two steps rather than one. First transcribe the voicemail in the language it was spoken β set the language manually rather than relying on auto-detect if the caller is hard to hear. Then translate the resulting transcript. Transcribing first is what makes the translation usable: a translation engine working from raw phone audio compounds two error sources, whereas a corrected transcript gives it clean input.
Do carriers offer voicemail to text? Most do, usually as a paid add-on to visual voicemail, and quality and availability vary by carrier, plan and country β check your carrier's own page before assuming you have it. On iPhone, Live Voicemail transcribes an incoming message on-device in real time as the caller speaks, but its language and region coverage is narrow; separately, whether a saved voicemail shows a transcript in the Phone app depends on your carrier's visual voicemail service.
Is there a voicemail to text app? Several, but for occasional use a browser is enough β you do not need to install anything to convert a file you already have. An app makes sense when you want every incoming voicemail transcribed automatically rather than converting messages one at a time. For confidential messages, a local app that transcribes on-device avoids uploading the audio anywhere at all.
How accurate is voicemail transcription? Accuracy percentages quoted by vendors are measured on their own audio and are not comparable to each other, so treat any single number with suspicion. What is predictable is where transcription fails: proper nouns, phone numbers read quickly, heavy background noise on the caller's end, two people talking, and callers switching language mid-message. Always check names and digits against the audio before acting on them.
Is it legal to transcribe a voicemail? A voicemail left for you is a message addressed to you, and transcribing it for your own use is ordinarily unproblematic β but recording laws, workplace policy and data-protection rules differ by country and by context, especially for business or client messages you intend to store or share. If the message is sensitive, transcribe it on-device rather than uploading it, and check your own obligations before republishing anyone's words.
How long can a voicemail be? Carriers typically cap individual messages at a few minutes, which is far below any transcription limit β a voicemail will not run into a file-size ceiling. TranscribeNext allows 50 MB and 2 hours per file on the free plan, 2 GB and 5 hours on PRO and 5 GB and 8 hours on BUSINESS, so the constraint is never the voicemail itself.
Sources
Related guides: Best Mac apps to transcribe audio Β· How to convert M4A to text Β· How to transcribe audio files quickly Β· Transcription tips for interviews
Convert your voicemail now
Save the message as a file and drop it into the form at the top of this page β the first 5 minutes are free with no account, which is longer than any voicemail your carrier will store. For messages you would rather not upload at all, TranscribeNext for Mac transcribes them on-device.