Auri Smart Notes: Voice, Transcripts, and AI Transforms

Two capture modes, one note

Smart Notes is built for people who sometimes type and sometimes talk, and who do not want those drafts living in three apps.

• Text. Markdown notes with photos, tables, and to-do lists in the same document, plus eight note colors for organization.
• Voice. Hit record, talk, get a transcript, then choose a transform. The transcript stays put; a transform is a view you can regenerate.

Keep the raw take when the wording matters. Keep the transform when the job is a summary, a checklist, or a first draft.

The transform menu, by job

After a transcript lands, pick the output that matches why you recorded. Do not run every option.

Here is what a transform does with a messy take:

You said: "okay so for the offsite, um, the room deposit is due Friday, I still owe Dana the headcount, and I need to book the projector, no wait, Dana is bringing the projector"

To-Do List typed: "Pay the room deposit by Friday. Send Dana the headcount. Projector: Dana is bringing it."

The filler is gone, the self-correction resolved to the version you meant, and the projector line records a fact instead of a task you no longer have. Run one transform, then check names and numbers by hand.

Import YouTube videos, audio files, and PDFs

When the source is already recorded, bring it in through the import button in Smart Notes, which takes three source types:

• YouTube Video. Copy the video link in YouTube, then paste or type it into the field (the prompt reads "Enter or paste a link to a YouTube video," and a Paste button fills it in one tap). Regular watch links, youtu.be links, and Shorts links all work. There is no share-sheet route from the YouTube app; the link is the way in.
• Audio/Video File. A voice memo from another app, a lecture recording, a video saved in Files.
• PDF/Image. A document, or a photo of one.

The transcript then takes the same transforms as a voice memo. Run Short Summary first to see whether the content deserves a deeper pass, and use Meeting Notes or Podcast Notes when you need structure rather than a rewrite. Imported minutes count against the same transcription allowance as recordings.

Engines, iCloud backup, and the free 30 minutes

Transcription for Smart Notes has its own engine setting, separate from the keyboard's: Settings > Transcription & Smart Notes > Recognition > Transcription Engine. The caption states its reach: "Used by recordings, imported audio, YouTube transcription, and Smart Notes." Five engines are on the list: On-Device, Legacy, Azura, Supernova, and Wizper.

On-Device is English-only, downloads about 460 MB once, and keeps transcription on the phone; when it cannot handle a job, the online engine you picked takes over. If a note will hold client names or medical detail, set the engine before you hit record. Changing it afterward does not rewind what already left the device. Recurring names go into Custom Vocabulary on the same Recognition screen; it works with Legacy, Azura, and Supernova, not with Wizper or On-Device, and the custom words guide covers building the list.

Storage and sync go through iCloud. The app puts it in one line: "Your recordings and transcripts are stored on your device and backed up on your iCloud account." A note's Share menu exports it as plain text, Markdown, RTF, PDF, HTML, or a Word document. The free plan includes 30 minutes of transcription in total, across recordings, imports, and YouTube; Pro removes the cap.

A workflow that sticks

1. Record the messy version while the thought is fresh. Do not narrate punctuation.
2. Transcribe immediately so you are not sitting on a pile of unlabeled audio.
3. Choose one transform that matches the next share: summary for yourself, meeting notes for the team, to-do list for tomorrow.
4. Fix names, dates, and numbers by hand.
5. Paste the result into the place people will actually read it.

Keyboard voice typing is still better for short replies in other apps. Smart Notes is for material you want to keep and reshape.

Mistakes that make transforms worse

When a summary comes back vague or a meeting note misses the decision, the recording is usually the problem rather than the transform.

• Recording three topics in one take. Transforms follow the structure you gave them. One subject per recording produces sharper output than a ten-minute ramble across four projects.
• Never saying the names. A meeting note cannot assign an owner you did not mention. Say who is doing what out loud, even if it feels stilted.
• Running three transforms on the same transcript. Pick the one that matches the next share. Stacking them gives you summaries of summaries, and detail drops each round.
• Editing the transform instead of the transcript. When a fact is wrong at the source, fix the note first, then run the transform again.
• Recording at arm's length. Bring the phone closer or use a headset. Distance costs more accuracy than any setting you can change afterwards.

Say a one-line header at the start of each recording: what this is, and who it is for. It takes three seconds and it improves every transform you run on it.

Where Smart Notes fits with the keyboard and Mac

Three Auri surfaces overlap here, and the difference is what happens to the text afterwards.

The failure mode is using Notes as a clipboard on the way to another app. When the text is going somewhere else immediately, dictate it there with the keyboard and skip the round trip.

FAQ