Microphone permission vs Speech Recognition permission on iPhone: what’s the difference?

Install a recorder app that also transcribes and you will usually see two separate permission prompts before you have recorded anything — one for the microphone, a second one, worded differently, for speech recognition. It is a reasonable thing to wonder about: why two, and does saying no to one quietly break the other?

Recording and transcribing are two different jobs

iOS treats capturing audio and interpreting audio as separate capabilities, handled by separate frameworks. Recording — turning sound into a stored file — goes through the system’s audio session APIs. Transcription — turning that sound into words — goes through Apple’s Speech framework, a distinct piece of the system with its own authorization step. An app can use either one without the other: something that only records never has to ask about speech recognition at all, and dictation in the keyboard uses speech recognition without ever saving a file.

What Microphone permission actually gates

This one is narrow: can the app turn the microphone on. Allow it, and pressing record works. Deny it, and the record button fails immediately — there is no audio to work with, so nothing downstream of it matters yet. This is the same permission every app that touches the microphone asks for, recorder or not, and it is unrelated to what happens to the audio afterward.

What Speech Recognition permission actually gates

This one is about a specific action: asking Apple’s speech engine, on-device model or server, to turn audio into text. It has its own row in Settings and its own authorization prompt, independent of the microphone. The detail that surprises people: this permission is required even when the on-device model handles everything and nothing is transmitted. "On-device" describes where the processing happens, not whether the system still wants you to have explicitly agreed to speech recognition running at all. Denying it does not touch recording — it only blocks the specific act of requesting a transcript.

Why iOS keeps the two apart

Because they are genuinely separate decisions for a person to make. Someone might be entirely comfortable with an app recording their voice for later playback while having no interest in that voice being turned into searchable, quotable text. Someone else might want dictation-style speech recognition in an app that never saves a recording at all. Bundling the two into one prompt would force a single answer onto two different questions. Splitting them means each shows up as its own toggle under Settings → Privacy & Security, and each can be granted or revoked on its own, at any time, without touching the other.

What actually changes depending on which one you allow

• Microphone allowed, Speech Recognition denied: recording works fully. Requesting a transcript for any recording fails until Speech Recognition is allowed.
• Microphone denied, Speech Recognition allowed: nothing to transcribe yet, because there is no way to capture audio in the first place. The app cannot record until Microphone is allowed.
• Both allowed: recording and transcription both work, and which path a transcript takes — on-device or Apple’s server as a fallback — is a separate question from whether the permission itself was granted.
• Revoking either one later, in Settings, only blocks that action going forward. Recordings and transcripts already produced stay exactly where they were.

How Voice Studio handles this

Voice Studio asks for Microphone to record and, separately, for Speech Recognition to transcribe, because it calls Apple’s Speech framework rather than shipping a model of its own. Transcription always attempts the on-device pass first and only retries with the on-device flag turned off, meaning the audio goes to Apple’s server, if that first attempt comes back empty or errors — so the Speech Recognition permission is doing real work on the on-device path too, not just gating a network call. There is no account and no server of Voice Studio’s own either way; recordings and transcripts stay in the app’s own storage on the device regardless of which permission you have granted.

Common questions

Does denying Speech Recognition stop me from recording?

No. Recording only needs Microphone permission. Speech Recognition is a separate permission that only gates asking for a transcript afterward.

Do I need to grant Speech Recognition even if I only want on-device transcription?

Yes. On-device recognition is still gated by the same system permission as the server path — where the processing happens is a separate question from whether transcription is allowed to run at all.

Can I revoke Microphone or Speech Recognition without affecting the other?

Yes. Settings → Privacy & Security lists them as separate rows, and each can be turned off independently at any time without touching the other.

If I revoke Speech Recognition later, are transcripts I already made deleted?

No. Revoking a permission only blocks that action going forward. Recordings and transcripts already produced stay in the app’s storage untouched.