This snapshot compares on device and cloud speech approaches for note capture.
Caption: On device is fast, private, and works offline. Cloud adds a network hop and varies with connection.
TL;DR
- On device: lowest latency, works offline, best for privacy.
- Cloud: may improve accuracy in some noise. Needs network and may send data.
- Brain Dump: on device first. Optional polish can be enabled later.
Latency snapshots
On iPhone 14 Pro (iOS 18, Oct 2025), text begins streaming in under 300 ms and continues near real time. Typical round trip cloud setups add 300 to 1200 ms before first characters appear, plus network variability.
Apple's SpeechAnalyzer framework now matches mid-tier Whisper models in accuracy while running entirely on-device. 9to5Mac testing showed it was 2x faster than MacWhisper running Large V3 Turbo.
A short story about flow
Ada journals while walking. She tried a cloud setup because a friend said it was smarter. The first words took half a second to show up. Her brain hit the brakes. She kept checking the screen. She forgot the third sentence.
She switched back to on device. Words appeared as she spoke. She finished in a minute and did not think about the app once. Flow won. The note got written.
Privacy
With on device transcription, audio stays on your device. Optional polish runs only if you switch it on. See our privacy policy and Key facts.
When to pick which
- On device: journaling, sensitive notes, low connectivity environments.
- Cloud: specialized vocab at the cost of connectivity and potential data transfer.
Accuracy in the real world
Cloud can help with rare jargon in some setups. But most accuracy losses come from wind, distance, or echo, not from the model. Fix the mic first. Try wired EarPods while moving. Keep the mic 15 to 25 cm from your mouth. See the mic guide.
FAQ
Is cloud always slower. Often, because it adds a network hop and server processing. Some setups stream partials fast, but jitter can still hurt flow.
Is on device always private. Yes for capture in Brain Dump. Optional polish is opt in.
Which is more accurate. It depends on noise and vocabulary. Mic technique helps more than model choice for most people.
Ready to try? Download on the App Store.