← Bryl
§ Glossary

What is offline transcription?

Offline transcription is speech-to-text that runs entirely on the device — your phone, tablet, or laptop — without uploading the audio to a server. A transcription tool that needs the internet to produce a transcript is online, regardless of whether the vendor describes it as “secure” or “HIPAA-compliant.” A tool that produces a transcript in Airplane Mode is offline.

How offline transcription works

Until recently, accurate transcription required server-class GPUs. Models like OpenAI's Whisper, Deepgram, and Speechmatics were big enough that running them on a phone was infeasible at acceptable speed and accuracy. That changed in 2024 and 2025. Two families of speech models now run faster than real-time on a recent iPhone or iPad:

  • Apple's Speech framework — the system-level on-device speech recognizer that ships with iOS, used by dictation and Siri. Accurate enough for note-taking; tightly integrated with the Apple stack.
  • Parakeet TDT v3 — NVIDIA's open-source automatic speech recognition (ASR) model, ported to Apple Silicon by the FluidAudio team. Runs at roughly 210x real-time on an iPhone 15 Pro and produces a Word Error Rate of about 6.4% on a 120-session talk-therapy corpus — within striking distance of cloud-grade Whisper.

Both models execute on the device's neural engine. The audio is processed in chunks, the model produces text, and the transcript appears — typically in seconds for a 50-minute session. No part of the audio crosses a network boundary.

How offline transcription differs from cloud transcription

Cloud transcription

A cloud transcription tool — Otter, Rev, Mentalyc, Upheal, Granola, Fireflies, Zoom transcripts — uploads the audio to the vendor's servers, runs a model there, and returns the transcript over the network. The audio sits on the vendor's infrastructure for some retention window. Privacy is managed by terms of service and, for healthcare, a Business Associate Agreement (BAA).

Offline (on-device) transcription

An offline transcription tool keeps the audio on the device for the entire process. The transcript is produced locally. There is no server in the loop, so there is nothing to subpoena, breach, or accidentally retain. Bryl is the offline option built specifically for therapy sessions.

Why offline transcription matters for therapy notes

Therapy sessions are some of the most sensitive content a clinician handles — trauma history, current suicidal ideation, custody-relevant disclosures, EMDR processing blocks. The natural question for any session-recording workflow is “what happens to this audio?”

With cloud transcription, the answer is always some variation of “it sits on a vendor's server, encrypted at rest, retained per their policy, accessible to their employees with appropriate role-based access, and recoverable under subpoena.” With offline transcription, the answer is “it never left your device.” The two answers carry different risk profiles for the patient and different liability profiles for the practice.

How to verify a transcription tool is actually offline

The honest test takes thirty seconds:

  1. Open the tool on a phone or tablet.
  2. Turn on Airplane Mode.
  3. Record a short test clip and tap Transcribe.
  4. If the transcript appears, the tool is offline. If it spins or errors, the tool was uploading the audio somewhere.

A more thorough verification involves connecting the device to a network proxy (Charles, Proxyman, Little Snitch, mitmproxy) and inspecting traffic during a recording. An offline transcription tool will produce zero outbound requests during the transcription step. What on-device AI actually means for therapy notes walks through that verification end-to-end.

Related terms