Turning a voice recording into text no longer needs an extra app on most current phones. On iPhone, the Voice Memos app transcribes speech on an iPhone 12 or newer, German included, either live while recording or afterwards, and "Copy Transcript" puts the whole text on the clipboard. On iPad it works from iPadOS 18.2, on Mac from macOS 15 on an Apple silicon machine, and the Notes app does the same thing while keeping audio and text together in one note. On Pixel phones the Recorder app transcribes while you record, in fifteen languages from the Pixel 6 onwards, German among them, and hands the result over as a .txt file or straight into Google Docs; on Galaxy devices that job belongs to Transcript assist inside the Voice Recorder. The difference that matters in practice isn't accuracy, it's what happens next: on the main route Apple gives you text on the clipboard, Google gives you a file. And for recorded conversations the bigger question is where the processing happens anyway, on the device or at the provider.
I record a lot while I'm out, in the car, between two appointments, walking in the evening. What used to go wrong for me was never the conversion into text, it was the filing. So today I decide where the text should end up before I hit record. How I take a workflow apart before automating it is on the Process optimisation page, and which tasks are suited to being handed to AI at all is covered in Which tasks you can hand to AI.
What iPhone, iPad and Mac do out of the box
Apple documents two apps that transcribe speech; they work very similarly.
In Voice Memos you start recording and swipe up on the waveform to watch the transcript run live. Afterwards you tap the recording and either choose "Show Transcript", select a passage and copy it, or take "Copy Transcript" for the whole thing. The search is genuinely useful: according to Apple the search field above the recordings list covers not just titles but the transcriptions too, and picking a hit moves the playback position to that spot.
In Notes you tap the plus inside a note and then "Record Audio". You open the transcript from the icon on the audio recording and either pull it into the note itself with "Add Transcript to Note" or copy it out. One detail from the manual: delete the audio recording and the transcript goes with it.
Apple states the requirements per device:
German is on the language list for all three, and for all three Apple notes that audio transcriptions aren't available in every country or region. What you actually see depends on model, system version, language and region at once, and the vendors keep changing that combination: every detail here is as of September 2026. A text-file export isn't described in the manual; copying into another document is. The audio file itself you export with "Save to Files", by default as .m4a. If all your devices share one Apple Account and "Voice Memos" is switched on in the iCloud settings, the recordings sit on all of them anyway.
What Android does out of the box
On Pixel phones you first pick a language in the Recorder app's settings, and from then on it transcribes while recording. Google names two tiers: the Pixel 3 through Pixel 5a transcribe live in eight languages, German included, and from the Pixel 6 it's fifteen. Re-transcribing afterwards covers more than forty languages. The export is the most convenient I know of on a phone: press and hold the recording, tap "Share", choose "Transcript (.txt)". The three-dot menu adds "Send transcript to Google Docs" and "Send transcript to NotebookLM". On storage, Google says recordings stay on the Pixel unless you trigger the sync yourself. Turning on backup makes that sync permanent; sharing, copying into another app or saving somewhere else such as Drive do something different and hand only that one recording or transcript to whichever service you pick in the moment.
On Galaxy devices the app is called Voice Recorder and the transcript feature belongs to Galaxy AI: open a recording, tap transcribe, pick the language, confirm. Samsung lists 23 languages and regional variants for it. The transcript can be translated and summarised, and for the summary Samsung requires you to sign in with a Samsung account. The audio files live in My Files under Internal storage > Recordings > Voice Recorder, and sharing works by pressing and holding and then tapping send. Samsung doesn't describe a separate text-file export on these help pages.
If your device is neither a Pixel nor a Galaxy, the route goes through an app. Which ones are worth looking at is in Transcription Software Compared.
Phone calls: what works and what doesn't
Phone calls come with stricter rules than a voice note.
On iPhone there is call recording, with the transcript available from iOS 18.1 according to Apple: during the call tap the further actions, choose call recording, and both sides hear an announcement. The recording lands in the "Call Recordings" folder in Notes, the transcript sits next to it and, Apple says, attributes every statement to the person speaking. Except: call recording isn't available in the European Union. Apple lists the EU explicitly among the regions without the feature. Next to it Apple puts the sentence I consider the most important on that page: verify the accuracy of a transcript before you rely on it, it may not be a perfect representation of the conversation.
On Android, the Google Phone app can record calls. Google describes the feature as available in many places but not in every country or region, and names Android 9 or newer on non-Pixel devices and a Pixel 6 or newer on Android 14 on Pixel devices. Both parties are informed when recording starts, and according to Google the recordings stay on the device and aren't backed up anywhere else. A transcript isn't part of it: you get an audio file, which you then feed into the Recorder app, the Voice Recorder or a transcription app. If you make your calls through Teams instead, the route is in Transcribing Calls in Microsoft Teams.
And the paragraph that comes before all of it: in Germany, recording another person's non-publicly spoken word without authorisation is a criminal offence under § 201 of the Criminal Code, not a mere administrative one. You are on safe ground once everyone else involved has explicitly agreed in advance. The phone's automatic announcement carries the information; the consent itself needs to be documented so you can prove it. So ask beforehand, record the agreement, and settle separately what may happen to the recording afterwards. The details, including the GDPR side, are in Transcribing Conversations: The Legal Ground Rules. I'm a developer, not a lawyer: this is my working basis, not legal advice.
Where the recording gets processed
For a shopping list it doesn't matter; for a client conversation it's the deciding question. So here is what the providers themselves state, not what could be inferred from it.
Two things follow for me. First: a statement missing from the page we evaluated is not a promise, and no evidence that the vendor makes none anywhere. "Not mentioned" means neither "stays on the device" nor the opposite. Second: it is precisely at summarising, translating and re-transcribing that Google and Samsung name the path outwards. If you want certainty that a recorded conversation never leaves the device, pick a tool that says so explicitly. Which data goes towards AI in my own setup is in AI and data protection: what the AI is allowed to see.
From the phone into your workflow
Three routes, ordered by effort. On Pixel you send the transcript straight to Google Docs, where it is searchable from any computer. On iPhone, iPad and Mac you paste the copied text into a destination you already search; with iCloud sync switched on you copy it on the Mac with a keyboard instead of a thumb. On Galaxy and everything else you share the audio file into a transcription tool: the detour, but also the route where you decide the processing location yourself.
The filing is the actual work
This is where most collections fall apart. Converting to text happens automatically now; the hard part is getting the text to land where you will find it again. Anyone who doesn't solve that ends up collecting audio files instead of knowledge.
What helps in my case are three decisions, all of them made before recording rather than after:
- One destination, not five. A single collection point you open daily anyway. Whether that's Notes, Obsidian, Notion or a folder in Drive matters less than the fact that there is exactly one.
- A name that says what's inside without listening. Date, topic, occasion. "Voice memo 47" is worthless in three weeks; "2026-09-19 quote renovation follow-up deadlines" isn't.
- A fixed slot for the transfer. Once a day, five minutes, everything from the phone into the destination. Without it, the best filing structure stays empty.
The test: if a week later you can't say whether an idea is in Voice Memos, in Notes or in a voice message to yourself, you don't have a transcription problem, you have a filing problem. A better app won't fix it.
Many voice notes become a register
One voice note becomes text. Many conversations become something else, and that's the step that pays off: with a stack of client conversations transcribed, what's in there is more than a row of separate details. It becomes visible which question keeps coming back and which of your own answers drew no follow-up question afterwards.
That is exactly what Gustav is built for, the AI employee on my team who evaluates call transcripts. He doesn't deliver minutes, he delivers a pattern register: the recurring questions together with your own best answer, which is where a playbook, an FAQ, an objection library and raw material for content come from. A local Whisper model transcribes the audio. Without a clean legal basis he never sees a transcript, and before any processing there is a fixed anonymisation pass in which names become roles, companies become industries, amounts become orders of magnitude and places become regions. Replacing names only makes the material pseudonymous, not anonymous (Art. 4(5) GDPR); the pass therefore includes a per-entry counter-check on whether role, industry, timing and a quotation together still point to one person. He also distils only what you said yourself, instead of inventing advice of his own. What that looks like step by step is in Conversation analysis with AI: the pattern register. If all you want for now is to follow up on individual meetings, Creating meeting minutes automatically is the closer starting point.
Frequently asked questions
Can an iPhone turn voice recordings into text without an extra app?
Yes. According to Apple, the Voice Memos app transcribes speech on an iPhone 12 or newer, German among the supported languages, live while recording or afterwards. The Notes app does the same for audio recorded inside a note. Apple points out that the feature isn't available in every country or region.
Why can't I find call recording on my iPhone?
Because it isn't offered in the European Union. Apple lists the EU explicitly among the regions where call recording is currently unavailable. In regions that have it, both sides hear an announcement.
How do I get a transcript as a file rather than text on screen?
Most directly on Pixel phones: press and hold the recording, tap "Share", choose "Transcript (.txt)", or alternatively "Send transcript to Google Docs". With Apple, the manual describes the clipboard route, so "Copy Transcript" and paste into a document. The audio file itself you export with "Save to Files".
Does the recording stay on the device or go to the provider?
It depends on the tool and on the feature. Google says re-transcribing may be processed on Google servers and that summaries run in an isolated cloud environment. Samsung states that Galaxy AI processes data on the device or via the cloud depending on the feature. Apple's manual pages name device requirements but no processing location.
What do I do with old voice memos that have no transcript?
Opening them is often enough. Apple says recordings from iOS 17 or earlier are transcribed automatically when opened, provided they contain speech; on the Mac the same applies to recordings from macOS 14 or earlier. In the Pixel Recorder app an existing recording can be re-transcribed afterwards, in more than forty languages according to Google.
Where to go from here
Take the next appointment where you'll want to capture something on the move anyway, and decide beforehand where the text should go. Repeat that daily for a week, then check whether you can find everything again. Only after that is it worth looking at tools, and that comparison is in Transcription Software Compared.
Once many recordings are supposed to become a register, the filing turns into a role. How I build such roles is collected under AI employees, and the route to your first one is in Hiring an AI employee. Templates and ready-made packages are in my community, Claude Practitioners.
