Echo Flow Get EchoFlow Free

← Back to Blog

Private Dictation · 8 min read

Where Your Dictated Words Go: A Packet Trace You Can Run Yourself

Where Your Dictated Words Go: A Packet Trace You Can Run Yourself

Where Your Dictated Words Go: A Packet Trace You Can Run Yourself

I wanted to know exactly where my dictated words go. So I ran a packet trace.

I'm a technical person. "Your data stays private" isn't enough for me , I want to see the proof. So I opened Little Snitch, started a dictation app, and watched what connections it made.

The results were educational. And a little scary.

1
Audio capture
Your microphone records audio. This happens locally , no network required. So far so good.
2
Transcription
Your audio becomes text. This can happen locally (Apple Speech framework) or in the cloud (server-side).
3
Polishing
Your raw text gets cleaned up. This is where most cloud processing happens , and where your drafts leave the device.
4
History
Your dictation gets stored. Locally (on your Mac) or in the cloud (on someone else's server).

The trace that changed my mind

I ran the trace on three popular dictation apps. Here's what I found:

App A (cloud dictation): Audio uploaded to a transcription service. Text sent to a separate AI service for polishing. History stored on their servers. Three different network connections for one dictation.

App B (hybrid): Audio transcribed locally. Text sent to cloud for polishing. History synced across devices. One network connection , but it carried my rough drafts.

App C (local): No network connections at all. Audio transcribed locally. Text polished locally. History stored locally. Zero packets sent.

App C is Echo Flow. I built it because I didn't like what I saw with Apps A and B.

Cloud dictation
Multiple network hops
  • Audio uploaded for transcription
  • Text sent for polishing
  • History stored on servers
  • Three data journeys per dictation
Echo Flow
Zero network required
  • Audio transcribed locally
  • Text polished locally
  • History stored locally
  • No packets sent

How to run this trace yourself

You don't have to take my word for it. Here's how to verify any dictation app's privacy claims:

Tool: Little Snitch (macOS) or LuLu (free, open source). Both show you every network connection an app makes.

Method:

  1. Install the firewall tool
  2. Open your dictation app
  3. Dictate something
  4. Watch what connections appear

If you see connections to transcription services, AI providers, or cloud storage , your audio or text is leaving your device. If you see nothing, it's processing locally.

I do this with every app I install. It takes five minutes and tells me more than any privacy policy.

0
Network connections made by Echo Flow during dictation , that's the architectural fact, not a policy promise

Why this matters more than you think

Here's the thing about dictation: it captures your unfiltered thoughts. The rough drafts. The half-formed ideas. The things you'd never send without editing.

If those drafts travel to a cloud server for polishing, then your most unfiltered writing is the writing most exposed.

I talked to a journalist who told me she stopped using a cloud dictation app after realizing her source quotes were being processed on someone else's hardware. "I never signed up to share my sources' words with a server farm," she said.

That stuck with me. Your dictation is often your most sensitive writing. It deserves the most protection.

The honest note

Local processing isn't magic. It's a choice. Apple's Speech framework makes local transcription possible. On-device models make local polishing possible. Echo Flow was built around those choices because I believe your drafts should stay yours , not because it's trendy, but because it's right.

The technical reality

Here's what "on-device" actually means technically:

  • Transcription: Apple's SFSpeechRecognizer runs on Apple Silicon Macs. No network required. Apple documents this.
  • Polishing: Local models on Apple Silicon (or Apple Intelligence on macOS 26) can run inference locally. No cloud API call.
  • History: Stored in a local database on your Mac. No sync, no server, no account.

This isn't new technology. Apple has been building local AI capabilities for years. Echo Flow just uses them.

✓
Run a packet trace , Little Snitch or LuLu, both free.
✓
Dictate something , watch what connections appear.
✓
Check the domains , transcription services, AI APIs, cloud storage?
✓
Verify the privacy policy , does it match what you saw?

The bottom line

"Privacy" is a word that gets thrown around without proof. I wanted proof. So I ran the trace.

If you dictate anything sensitive , client work, medical notes, legal strategy, journalism , run the trace yourself. Five minutes tells you more than any marketing copy.

I built Echo Flow to pass that trace. Zero connections. Zero uploads. Zero doubts.

Try Echo Flow free for 14 days · Why on-device dictation · How local polishing works

Sources