Key takeaways
- •M4A is iPhone Voice Memos' default format — AAC audio in MP4 container, ~30-50 MB/hr.
- •AI transcribes M4A directly — no format conversion step needed.
- •We publish no accuracy percentage, and explain why below; review proper nouns and technical terms.
- •Free options: 5 minutes with no account at all, 30 minutes of credit with a free one — and on an iPhone 12 or later, your phone may already do it.
- •Output formats: TXT, DOCX, JSON, SRT from a single transcription pass.
- •Cost $0.20-$0.60 per audio hour AI; paid plans start at $2/month.
- •Processing 3-5 minutes per audio hour AI; 12-48 hours human turnaround.
- •Diarization (speaker labels) included on every paid plan — useful for interviews.
Doing it for free
Several options cost nothing, and the honest answer depends on how long the recording is and what you need back. Checked against each provider's own page on 18 September 2026:
| Option | Free tier | Account? | The catch |
|---|---|---|---|
| VexaScribe | First 5 minutes of any file up to 200 MB, no account. 30 min credit with a free account, files to 5 GB. | No, for the preview | The free credit is one-time, not monthly. |
| Apple Voice Memos | Unlimited, on-device, no upload at all | No | iPhone 12 or later only, and about ten languages. Getting the transcript back out of the app is awkward. |
| AudioScribe | 5 minutes / 100 MB, or 25-minute files with a free account | No, for the short tier | Publishes no accuracy figure at all, which is more honest than most. |
| Any2Text | First 15 minutes | No | Claims “up to 98% accuracy” with no source. |
| Notta | One file, four minutes maximum | Email required for the result | Its MP4 page advertises five minutes for the same engine, and it quotes “up to 98.86%” with no source. |
Self-hosted Whisper belongs on this list too if you are comfortable with a terminal: it is free permanently, handles .m4a natively with no extraction step, and nothing leaves your machine. It needs a GPU to be quick and some setup.
Free tiers move around. Check the provider's page before building a workflow that depends on one.
Convert a single .m4a or bulk-transcribe multiple .m4a files
VexaScribe handles both workflows from the same interface. Drop one .m4a file for a single conversion, or drop several files together several .m4a files for parallel bulk transcription.
Single .m4a workflow
Drag one .m4a file to the uploader (or paste an audio URL). Best for one-off iPhone Voice Memo transcription, a single interview, a single lecture recording. Free 30-minute trial covers the first single file end-to-end with all four export formats. Typical .m4a from a 30-minute Voice Memo processes in 3-8 minutes.
Bulk .m4a workflow
Select or drag several .m4a files at once. Each transcribes independently — you do not wait for the slowest. Useful for a semester of lecture recordings, a podcast back-catalog (episodes exported as .m4a from GarageBand/Logic), an interview series (research studies with 20-40 subjects), or a music-teacher's student recital archive. Each file counts against your monthly minutes.
Every uploaded .m4a file gets the same treatment — transcription, automatic speaker diarization (renamable in the editor), word-level timestamps, and language auto-detection across 99 languages. Mixed-language batches are fine — each file's source language is detected independently.
High-volume research or archives. Teams processing 50+ M4A recordings monthly (interview studies, oral history archives, dictaphone libraries) benefit from flat-rate M4A batch processing rather than per-minute charges — up to 6,000 minutes/month covered under a single subscription.
How to convert M4A to text (4 steps)
- 1
Upload the M4A file
VexaScribe accepts M4A directly up to 5 GB per file (approximately 16-20 hours of typical iPhone Voice Memos). No format conversion to MP3 or WAV required. Free trial accepts the first 30 minutes — covers a typical short iPhone Voice Memo end-to-end.
- 2
Choose source language and diarization
Select source language from 99 supported languages, or use auto-detect for clean monolingual audio. Toggle speaker diarization on for interviews, multi-speaker recordings, or seminar discussions. Diarization is included on every paid plan with no tier gating.
- 3
Wait for processing
A 30-60 minute M4A finishes in roughly 2 to 5 minutes on the standard model — medians measured across 7,432 production jobs. VexaScribe emails you when the transcript is ready. While waiting, queue additional M4A uploads — useful for batch transcription across an interview series or podcast archive.
- 4
Download the transcript
Pick the output format that fits your downstream workflow: TXT (plain text), DOCX (formatted with timestamps and speaker labels), JSON (structured for developer pipelines), or SRT (subtitle file when the M4A is the audio track of a video). All four formats export from a single transcription pass.
Getting M4A off your iPhone
Voice Memos saves recordings as .m4a in the app, but the file lives inside the iPhone's app sandbox — you need to export it before uploading to a transcription tool. Three methods cover virtually all workflows.
1. AirDrop to Mac or iPad (fastest)
Open Voice Memos → select the recording → tap the share icon → AirDrop → pick your Mac or iPad. The .m4a transfers in seconds. From the Mac, upload directly to VexaScribe. Best for: same-room transfer, large files.
2. iCloud Drive (multi-device sync)
Voice Memos → share → Save to Files → choose iCloud Drive. The .m4a syncs to any device signed into your Apple ID. Best for: workflows where you record on iPhone but work on a desktop later, or sharing between devices not in the same room.
3. Email to yourself (universal fallback)
Voice Memos → share → Mail → send to your own email address. Works on any platform — open the email on Windows, Linux, or Android and download the attachment. Best for: cross-platform workflows. Caveat: email attachment limits are typically ~25 MB, which caps you at roughly 50-60 minutes of M4A; for longer recordings, use AirDrop or iCloud.
File size reference. iPhone Voice Memos at default 64 kbps AAC produces roughly 30 MB per hour of recording. Lossless ALAC (iOS settings → Voice Memos → Audio Quality → Lossless) produces 5-10× larger files but maximum-quality source for transcription. For interview recordings under 4 hours, default quality is fine for transcription; for long lectures or critical archive recordings, Lossless is worth the storage cost.
Your iPhone might already do this
Since iOS 18, Voice Memos transcribes recordings on the phone itself. Open the recording and tap the transcript icon — nothing uploads anywhere, and it costs nothing. If that covers your situation, use it. It is genuinely the better option and you can stop reading here.
Use the built-in feature when
- ●You are on an iPhone 12 or later — it runs on-device and needs the A14 chip, so older phones cannot get it through a software update.
- ●The recording is in one of roughly ten supported languages: English, Spanish, Portuguese, Italian, French, German, Japanese, Korean and Chinese.
- ●The audio cannot leave your phone — a source under confidentiality, a privileged conversation, anything you would not upload.
- ●You just need to read it, and reading it inside the app is fine.
It falls short when
- ●Your language is outside that set. Record in an unsupported language and the transcript comes back empty or as a rough phonetic approximation.
- ●You need speaker labels. It produces one continuous block of text, which is a problem for any interview or meeting.
- ●You need the text as a file. Getting the transcript out of the app is awkward, and there is no DOCX, SRT or JSON export.
- ●The recording did not come from Voice Memos — a GarageBand bounce or a podcast download is not covered.
In short: for a short English memo you only need to read, the phone wins. For an interview you need attributed, exportable and searchable, it does not.
What is an M4A file?
M4A (MPEG-4 Audio) is an audio-only file format using the MP4 container with an AAC audio codec. File extension .m4a; MIME type audio/mp4. Apple introduced M4A as the default for iTunes audio and Voice Memos because AAC delivers better quality than MP3 at the same bitrate.
What matters for transcription
- →AAC bitrate. ≥64 kbps recommended for transcription. iPhone Voice Memos default is 64 kbps AAC — fine for most use cases. 32 kbps or lower degrades accuracy noticeably.
- →Lossy compression. Unlike WAV (uncompressed PCM), M4A is lossy. For studio-grade source where every dB matters, use WAV instead — see WAV to text.
- →ALAC variant. iOS "Lossless" Voice Memos setting uses ALAC (Apple Lossless Audio Codec) inside the M4A container. Larger files, no quality loss vs the source — best transcription accuracy possible from an M4A.
- →M4P warning. M4P is M4A with FairPlay DRM (Apple Music purchases). Transcription tools can't decode protected audio. Voice Memos and personal recordings are never .m4p.
- →Container kinship with MP4. M4A uses the same MP4 container as video files — that's why some tools treat M4A and MP4 interchangeably. See MP4 to text for the video-container sibling page.
Output formats (TXT, DOCX, JSON, SRT)
Four output formats cover most downstream workflows. VexaScribe exports all four from a single M4A transcription — no re-processing required.
| Format | Best for | Notes |
|---|---|---|
| TXT | Quick reference, copy-paste into Apple Notes, Word, Google Docs | Plain text — no timestamps, no speaker labels |
| DOCX | Editing in Word, sharing with stakeholders, deliverables | Formatted Word document with timestamps and speaker labels |
| JSON | Developer workflows, structured pipelines, custom integrations | Word-level timestamps and speaker IDs, machine-readable |
| SRT / VTT | When the M4A is the audio track of a video, or for a captioned podcast | Timestamped subtitle files, UTF-8 encoded |
| MP3 | Handing the audio to someone whose software will not open .m4a | A more universally accepted container than M4A |
Picking the right format. If you're reading the transcript yourself or pasting into Apple Notes, TXT is fine. For sharing with stakeholders, DOCX preserves structure. For building a search index or custom integration, JSON gives word-level timestamps. SRT is mainly useful if the M4A is the audio track of a video (e.g., podcast episode with a YouTube video version) — see video to SRT for the dedicated subtitle workflow.
About “98.86% accuracy”
Nearly every tool in this category publishes an accuracy figure, and almost none says what audio it was measured on or against what reference transcript. One result on the first page of this search advertises “up to 98.86% accuracy” for “high-quality audio” that is never defined, with no source. The same provider caps this tool's free tier at four minutes while its MP4 page says five — same company, same engine, a different number on each landing page. That is worth noticing, because it shows these figures are set by whoever writes the page rather than by anything measured.
So we do not publish one either. We have not measured our own word error rate against reference transcripts, and quoting a number without having done that would be the same marketing we are describing. What we can tell you is what actually shifts the result on a Voice Memo — and it matters far more than which tool you choose.
- 1
Where the phone was
This dominates everything else. A phone on the desk between two people and a phone in a bag produce completely different transcripts from identical software. If you can change one thing about how you record, change this one.
- 2
Whether people talk over each other
The hardest case for any speaker separation, ours included. For an interview you can plan ahead: record on two phones, one per speaker, transcribe each separately, and the attribution is near-perfect.
- 3
The recording-quality setting
Voice Memos defaults to 64 kbps AAC, which is adequate. The Lossless setting produces much larger files but keeps everything the microphone heard. For an interview that will be quoted, it is worth the storage.
- 4
Names, jargon and acronyms
Proper nouns are where corrections concentrate, because they cannot be worked out from context. A targeted search-and-fix pass beats reading the whole transcript.
- 5
Distance and room
A lecture recorded from the middle of a hall is a harder problem than the same lecture recorded at the front, and no software closes that gap afterwards.
The honest way to decide: run the first five minutes of your own recording through the box at the top of this page, with no account, and look at what comes back. That tells you more about your file than any percentage — including one of ours. For anything published or legally consequential, budget a human read regardless of the tool.
What you can do once you have the transcript
Voice Memos are usually recordings of something — an interview, a lecture, a meeting, a thought you wanted to keep. The transcript is rarely the end of the job. Four things run on the text afterwards without re-uploading anything:
Six summary types, not one
General, Meeting, Sales call, Interview, Lecture and Podcast. The type changes what gets extracted — an Interview summary looks for what the subject actually said, a Lecture summary follows the argument. For a Voice Memo of a research interview, Interview is the one you want.
Ask the recording a question
Rather than scrubbing a two-hour memo, ask when a particular topic came up and get the answer with the timestamp it came from, so you can jump to that point and check it yourself.
Translate without re-uploading
Send the finished transcript into another language with speaker labels and timings preserved. Useful when the recording is in one language and the people who need it read another.
Six export formats from one run
TXT, DOCX, SRT, VTT, JSON and MP3. One processing run produces all of them, so you do not re-upload to change your mind about the format.
Before you publish: four things to check
Every automatic transcript needs a read. These four spots are where corrections cluster, so checking them specifically is faster than reading top to bottom.
- 1
Names, companies and acronyms
The most frequent correction, because a name cannot be guessed from context. Use the editor's search: once for the correct spelling, once for the likely mis-hearing. On a recording where the same names recur, this saves the most time.
- 2
Numbers, dates and figures
The same document can contain both “twenty twenty-six” and “2026”. If the transcript is going into a report or notes, normalise these in one pass first.
- 3
Homophones
their/there/they're, its/it's. These slip past because the sentence still reads correctly. Searching for the usual suspects is faster than proofreading.
- 4
The speaker handovers
Speaker separation is least certain where two people overlap. Jump between speaker changes, check the first line after each, and rename the speakers while you are there — the rest of the review is quicker once the labels are real names.
If you are quoting someone verbatim — in an article, a dissertation, or notes that will be signed off — click the quote in the editor and listen back. Every word carries a timestamp, so it takes seconds, and it is the only way to be certain about a quote that goes public.
Cost: per-M4A and bulk math
M4A transcription is cheap on AI tools — typically $0.20-$0.60 per audio hour. Apple Voice Memos' built-in iOS 18+ transcription is genuinely free (on-device). Human transcription runs 150-1,500× more expensive than AI and is only justified for court-grade verbatim.
| Tool | Per audio hour | Entry plan | Best for |
|---|---|---|---|
| VexaScribe | $0.20-$0.60 | $2/mo, or 5 min free with no account | Most M4A transcription — speaker labels, timestamps, six export formats |
| Apple Voice Memos (iOS 18+) | $0 | Built-in | On-device, iPhone 12 or later, about ten languages |
| Rev AI | ~$6/hr ($0.10/min) | PAYG | Developer/API integration |
| Self-hosted Whisper | $0 forever | n/a | Technical users with a GPU |
| Human (Rev, 3PlayMedia) | $90-$300/hr | per-minute | Court-grade, verbatim, broadcast-certified |
Bulk math example. A grad student archiving weekly 60-minute interviews = 4 hours of M4A per month. AI transcription costs $0.80-$2.40 vs $360-$1,200 for human. A semester of 50 lectures × 50 minutes ≈ 42 hours of M4A total — $8-$25 on AI vs $3,780-$12,600 human. For full cost analysis, see how much does transcription cost?.
M4A files that are not in English
The language is detected from the audio, so there is nothing to set before uploading, and a recording that moves between two languages is handled without you picking one. This is the main practical gap versus Apple's built-in transcription, which covers about ten languages. For which languages are well covered and which need more editing, the main audio-to-text page goes through that.
Common M4A errors and fixes
Most M4A transcription problems come from one of five issues. Here's how to recognize and fix each.
Upload rejected — file is .m4p not .m4a
Cause. M4P is M4A with FairPlay DRM — Apple Music purchases and rentals download with this extension. Transcription tools can't decode protected audio.
Fix. If the source is Voice Memos or unprotected audio, the file should already be .m4a. If you have a legitimate need to transcribe a protected Apple Music file, you'll need to play it back through audio capture (with permission) — direct decoding isn't possible without removing DRM, which has licensing implications.
File size over 5 GB
Cause. Long-form recordings — a multi-hour conference recording or podcast archive M4A can exceed VexaScribe's 5 GB per-file limit.
Fix. Split with QuickTime Player on Mac (Edit → Trim, save segments). Or upload segments separately and concatenate transcripts at the end. The 5 GB limit covers approximately 16-20 hours of typical iPhone Voice Memo at 64 kbps AAC.
Corrupted M4A from interrupted iCloud sync
Cause. Voice Memo that didn't finish syncing through iCloud may have a truncated or corrupted file structure. Common when iCloud storage fills up mid-sync.
Fix. Re-export from Voice Memos to a different location (AirDrop or local Files). If that fails, re-mux with ffmpeg: ffmpeg -i broken.m4a -c copy fixed.m4a — often rebuilds the container without re-encoding the audio.
Speaker labels mixed up on interview recordings
Cause. Diarization struggles when both speakers share one microphone with similar voices, or when one speaker is much closer to the iPhone than the other.
Fix. If you have the chance to re-record an interview: use two iPhones, one per speaker, transcribe each separately, then interleave by timestamp. The result is near-perfect speaker attribution. For already-recorded mixed-channel M4As, manually re-label speakers in the DOCX export.
Accuracy unexpectedly low
Cause. Voice Memo recorded with iPhone in pocket / bag (muffled), at distance (>2 meters from speaker), or outdoors with wind. Audio capture conditions matter more than the file format.
Fix. Re-record closer to the speaker if possible. For existing low-quality M4As, budget extra review time, and consider pairing with human transcription for critical content.
M4A to text vs alternatives
We position VexaScribe honestly: it's the right pick for most M4A transcription needs across the Apple ecosystem (direct upload, 99 languages, speaker labels, multi-format export, $2/mo entry). Other tools win specific lanes — here's the honest read.
| Tool | Best for | Entry price | Direct M4A? |
|---|---|---|---|
| VexaScribe | Most M4A transcription — speaker labels, timestamps, six export formats, typed summaries | $2/mo, or 5 min free with no account | Yes |
| Apple Voice Memos | Short English memos on iPhone 12 or later, nothing leaves the phone | $0 | Yes (in-app only) |
| AudioScribe | A one-off short file with no account | $0 | Yes |
| Descript | Editors working the M4A into a podcast or video | $16/mo (10 hrs) | Yes |
| Self-hosted Whisper | Technical users, high volume, free forever | $0 | Yes (native, no extraction step) |
When to pick something other than VexaScribe. If you have a recent iPhone with iOS 18+ and only need English transcription with maximum privacy, Apple's built-in Voice Memos transcription is genuinely good and free. If you're editing podcast video and want the transcript inside the same tool, Descript fits. If you have technical skills, a GPU, and high volume, self-hosted Whisper is free forever — pay the setup cost once, run unlimited. For court-grade verbatim or broadcast-certified output, human transcription is the right call.
See also transcription tool alternatives and the AI vs human decision framework.
FAQ
Frequently Asked Questions
How do I convert M4A to text?
Four steps. (1) Upload the .m4a file to an AI transcription tool — VexaScribe accepts M4A directly up to 5 GB, no format conversion step needed. (2) Choose source language (auto-detect works for clean monolingual audio) and toggle speaker diarization on for interviews or multi-speaker recordings. (3) Wait roughly 3 to 5 minutes per audio hour. (4) Download the transcript in TXT (plain text), DOCX (formatted with timestamps and speaker labels), JSON (structured), or SRT (subtitle file). Total time from upload to ready transcript: 10-25 minutes for a typical 30-60 minute M4A.
Can I convert M4A to text for free?
Yes, several ways — all checked on 18 September 2026. If the recording is a Voice Memo in one of about ten languages and you are on an iPhone 12 or later, your phone transcribes it on-device for nothing and the audio never leaves the handset; that is genuinely the best option when it applies. Otherwise: the first five minutes of any file up to 200 MB are free here with no account, and a free account adds 30 minutes of credit. AudioScribe does five minutes with no account, or 25-minute files with one. Notta allows a single file of four minutes. Self-hosted Whisper is free permanently, handles .m4a natively with no conversion step and keeps everything on your machine, but it needs a GPU to be quick and some setup.
How do I get an iPhone Voice Memo as text?
Three transfer methods then transcription. (1) AirDrop to a Mac or iPad — Voice Memos app → select recording → share icon → AirDrop → receive on the other device. Fastest method. (2) iCloud Drive — Voice Memos → share → Save to Files → iCloud Drive folder. Accessible from any device signed into your Apple ID. (3) Email to yourself — Voice Memos → share → Mail → send to your own address; works on any platform but limited to ~25 MB attachment size. Once the .m4a is on a computer, upload it to VexaScribe (free the first 5 minutes are free with no account, and a free account adds 30 minutes of credit) or any M4A-supporting transcription tool. On iOS 18+, Apple Voice Memos also has built-in on-device transcription — tap the transcript icon directly in the recording, no upload needed.
What's the best M4A to text converter?
Depends on use case. For one-off short M4As (single iPhone Voice Memo): VexaScribe 30-min free trial — full feature access, no card, all four export formats. For iOS 18+ users wanting on-device privacy: Apple Voice Memos built-in transcription is genuinely good for English. For ongoing M4A transcription with 99-language support, speaker labels, and AI summaries: VexaScribe paid plans at $2-$20/mo. For technical users with privacy-critical content: self-hosted Whisper, free forever. For developer API integration: Rev AI at $0.10/min PAYG. Most Apple ecosystem users land on either Apple's built-in transcription (for quick English) or VexaScribe (for everything else).
How accurate is M4A transcription?
We do not publish a percentage, and it is worth saying why. We have not measured our own word error rate against reference transcripts, and a figure without a corpus and a method says very little — the same system behaves very differently on a phone sitting between two people and a phone in a bag. What actually moves the result, in rough order: where the phone was, whether people talk over each other, the recording-quality setting you used, how many names and acronyms appear, and the distance to the speaker. Several tools on this search advertise figures like "up to 98.86%" with no source — one of them caps its free tier at four minutes here and five on its MP4 page for the same engine, which shows how these numbers get set. The honest test is to run the first five minutes of your own file, free and without an account, and look at what comes back.
Can I transcribe M4A in languages other than English?
Yes — VexaScribe supports M4A transcription in 99 languages via Whisper Large-v3, including Spanish, French, German, Italian, Portuguese, Japanese, Korean, Mandarin, Arabic, Russian, Hindi, Turkish, plus 87 more. Source language is auto-detected or manually selectable. Translation to 133 target languages is included on every paid plan — useful for ESL students transcribing English lectures and translating to native language, journalists recording field interviews abroad, multilingual researchers, and Apple ecosystem users outside the English-speaking world. Apple's built-in Voice Memos transcription supports a much narrower set of languages (about ten languages as of iOS 18) — for languages outside Apple's set, AI tools like VexaScribe are the realistic option.
Why does my M4A fail to upload?
Three common causes. (1) DRM-protected M4P — Apple Music purchases and rentals download as .m4p (M4A with FairPlay DRM); transcription tools can't decode these. Fix: in older iTunes you could burn to CD then re-rip; in current Apple Music, the protected file isn't transcribable without removing DRM (which has licensing implications). For Voice Memos or unprotected audio, this doesn't apply. (2) File size over 5 GB — VexaScribe's per-file limit covers approximately 16-20 hours of M4A at typical bitrates. Fix: split with QuickTime Player on Mac (Edit → Trim) or upload segments separately. (3) Corrupted M4A — interrupted iCloud sync or partial download. Fix: re-export from Voice Memos to a different location, or re-mux with ffmpeg (ffmpeg -i broken.m4a -c copy fixed.m4a).
Should I convert M4A to MP3 first?
No. Upload the M4A directly. Modern AI transcription tools (VexaScribe, Descript, Otter, Rev, Whisper-based services) accept M4A natively — no conversion step needed. Pre-converting M4A to MP3 adds a lossy-to-lossy re-encoding pass that slightly degrades audio quality and can marginally reduce transcription accuracy. The only case where conversion helps is if your transcription tool genuinely doesn't support M4A (rare in 2026). For self-hosted Whisper, M4A is supported natively — no ffmpeg pre-extraction needed for audio-only M4A files (unlike MP4 video which needs audio extraction first).
How long does it take to transcribe a 1-hour M4A?
Roughly 3 to 5 minutes of processing per audio hour on the standard model, or 1 to 2 minutes on premium — medians measured across 7,432 production jobs. Add review time on top: plan a pass for names, numbers and the speaker handovers rather than a full read. A 30-60 minute Voice Memo is typically done in two to five minutes.
What output formats can I get from an M4A transcription?
Four formats covering most workflows. TXT (.txt) — plain text, no timestamps, copy-paste into Notes, Word, Google Docs. DOCX (.docx) — formatted Word document with timestamps and speaker labels for sharing with stakeholders. JSON (.json) — structured output with word-level timestamps and speaker IDs, for developer pipelines and custom integrations. SRT (.srt) — UTF-8 timestamped subtitle file, used when the M4A is the audio track of a video (e.g., podcast episode with video version). VexaScribe exports all four formats from a single M4A transcription pass — no re-processing required.
Can I convert multiple .m4a files at once (bulk)?
Yes. You can select several .m4a files at once and each transcribes independently, finishing at its own pace rather than waiting on the slowest. Useful for a semester of recorded lectures or an interview series. Each file counts against your monthly minutes, and speaker separation, language detection and every export format run per file, so a batch with more than one language is fine.
How is .m4a different from .mp3 for transcription accuracy?
Effectively no difference for transcription. Both formats are lossy compression, and Whisper Large-v3 (the model VexaScribe uses) resamples both to 16 kHz mono internally before transcription — the format identity is lost before the AI ever sees it. What matters is source recording quality, not container format. An iPhone Voice Memo M4A at 64 kbps AAC and a podcast MP3 at 128 kbps CBR produce equivalent transcription accuracy on the same audio. Don't waste time converting M4A to MP3 to "improve" accuracy — the re-encoding actually adds tiny generation loss without benefit.
When another tool fits better
We are built for one job: you have a recording and you need the text, with speakers marked and the ability to export it. Four situations call for something else.
A short English memo on a recent iPhone
Use the built-in Voice Memos transcription. It is free, it is instant, and the audio never leaves the phone. We are the better answer once you need speaker labels, a file you can send someone, or a language Apple does not cover.
One short file and you would rather not sign up at all
Our free credit is one-time rather than monthly. AudioScribe, also on the first page of this search, does five minutes with no account and 25-minute files with a free one, and notably publishes no accuracy figure at all — which is more honest than most of this category manages.
The recording must never leave your machine
For a source under confidentiality, a privileged conversation, or anything covered by an NDA, no cloud tool is the right answer — ours included. Self-hosted Whisper handles .m4a natively and keeps everything local. Apple's on-device transcription does too, within its language limits.
You need a certified transcript
Court proceedings, regulatory filings and anything requiring an attestation of accuracy need a certified human transcript. No automatic tool provides that.