Hush Echo

Confidential audio transcription where your voice never leaves hardware we own.

GPU-accelerated speech-to-text transcription on our own hardware. We don't save your audio, it is encrypted in transit (TLS) and at rest (Fernet), processed, then secure-deleted on completion. No US cloud. No CLOUD Act. No compromise.

Start Transcribing →
9 audio formats. 5 output formats. Zero data retention.
Fernet encrypted
We own the hardware
Secure-deleted on completion
ICO ZC126901

Cloud transcription sends your voice to US servers. Ours does not.

Every major transcription service processes audio on US-controlled infrastructure. The CLOUD Act gives US authorities the legal power to compel access to that data, regardless of where you are.

Cloud Transcription

  • Audio uploaded to US-owned servers
  • Subject to CLOUD Act (H.R.4943)
  • May retain audio for model training
  • Patient and client confidentiality at risk
  • No control over data lifecycle

Hush Echo

  • Audio stays on hardware we own
  • Structurally outside the CLOUD Act
  • Encrypted in transit (TLS) and at rest (Fernet)
  • Secure-deleted immediately on completion
  • Full audit trail for compliance

Enterprise-grade transcription. Hush by design.

State-of-the-art speech-to-text running on GPU-accelerated hardware we own. No API calls to OpenAI, Google, or any US provider.

State-of-the-Art Accuracy

Best-in-class speech recognition, running locally on GPU-accelerated hardware for fast, precise transcription.

GPU-ACCELERATED

Auto Language Detection

Automatically detects the spoken language, or specify it manually. Supports 99+ languages with high accuracy.

99+ LANGUAGES

Fernet Encryption

Audio is Fernet-encrypted the moment it arrives. Decrypted only during processing. Secure-deleted immediately on completion.

ZERO AUDIO RETENTION

Queue System

Large files over 10MB are queued and processed asynchronously. You receive a job ID and get notified when transcription completes.

FILES >10MB SUPPORTED

REST API

Full REST API for programmatic access. Submit audio, check status, retrieve transcripts. Integrate into any workflow.

API ACCESS

MCP Protocol

Also available via the Model Context Protocol for direct integration with AI assistants and agentic workflows.

MCP COMPATIBLE

9 audio formats in. 5 output formats out.

Upload in whatever format you have. Export in whatever format you need.

.MP3MPEG Audio
.WAVWaveform
.M4AMPEG-4 Audio
.OGGOgg Vorbis
.FLACLossless
.WEBMWebM Audio
.WMAWindows Media
.AACAdvanced Audio
.MP4MPEG-4 Video
Plain Text.txt
SRTSubtitles
VTTWeb Subtitles
JSONStructured
Verbose JSONTimestamped

Transcription for every industry that handles sensitive audio.

When confidentiality is not optional, sovereign transcription is the only answer.

Medical Dictation

GP consultations, ward rounds, clinical notes. Transcribe sensitive patient interactions without sending audio to US servers.

GDPR · CONFIDENTIAL

Legal Proceedings

Depositions, witness statements, police-station (PACE) interview recordings. Legal privilege preserved on sovereign hardware.

LEGAL PRIVILEGE · SECURE

Meeting Minutes

Board meetings, strategy sessions, governance calls. Transcribe sensitive corporate discussions without third-party exposure.

CORPORATE · GOVERNANCE

Interview Transcription

HR interviews, journalistic sources, research interviews. Protect interviewee identity and sensitive content.

HR · JOURNALISM · RESEARCH

Podcast & Content

Podcast episodes, video content, lectures. Get accurate transcripts for show notes, captions, and accessibility.

CONTENT · ACCESSIBILITY

Insurance Claims

Claim calls, adjuster recordings, dispute resolution audio. Maintain chain of custody on UK infrastructure.

INSURANCE · COMPLIANCE

Upload. Transcribe. Delete. Your audio is gone when the transcript is ready.

Your audio is encrypted the moment it arrives and secure-deleted the moment transcription completes.

01

Upload

Submit your audio file via the web interface, REST API, or MCP protocol. Fernet-encrypted on arrival.

02

Transcribe

Your audio is transcribed on GPU-accelerated hardware we own. Language auto-detected or specified.

03

Retrieve

Download your transcript in your chosen format: plain text, SRT, VTT, JSON, or verbose JSON with timestamps.

04

Secure Delete

Audio and transcript are securely deleted from our hardware after processing. No audio or transcript retained.

Before your compliance officer asks.

Here is what is already true today.

🔒

Our Hardware, Our Rules

Audio is processed on hardware we own, then wiped. No rented cloud, no third party. No audio or transcript kept on our servers.

🔒

Fernet Encryption

Symmetric authenticated encryption. Audio encrypted in transit and at rest during processing.

🚫

Zero Audio Retention

Audio and transcripts secure-deleted on completion. No audio stored, no training, no audio retained.

📜

Full Audit Trail

Job metadata logged for compliance. File hashes, timestamps, completion status. Exportable anytime.

🛡️

ICO Registered

Data controller ZC126901. GDPR and Data Protection Act 2018 compliant.

🖥️

Our own hardware, not Big Tech

Hardware we own. No US parent. No US jurisdiction. Structurally separate.

Your voice deserves hardware we own.

GPU-accelerated transcription on hardware we own. Encrypted in transit. Deleted on completion. No cloud. No compromise.

Start Transcribing →

Or email [email protected] for enterprise pricing.

Hush Echo is an automated transcription tool. Always verify transcripts for accuracy before relying on them for clinical, legal, or regulatory purposes.

Frequently asked questions

Is Hush Echo a confidential transcription service?

Echo is a self-serve confidential transcription tool rather than a human transcription service: state-of-the-art speech-to-text on GPU hardware we own, with your audio encrypted in transit and at rest, then secure-deleted on completion. If you need certified human transcripts for court, use a human service; if you need fast, private working transcripts, that is Echo.

Can solicitors transcribe police-station (PACE) interview recordings with it?

If you hold a recording of a PACE interview or client meeting, Echo will transcribe it without the audio ever leaving hardware we own, and the file is secure-deleted after processing. Echo does not produce certified evidential transcripts, and nothing here is advice on admissibility; always verify the transcript against the recording.

Is this a secure transcription option for sensitive recordings?

Audio is encrypted in transit (TLS) and at rest (Fernet), processed on hardware we own, then secure-deleted, and it is never used to train models. Hush AI Limited is an ICO-registered data controller (ZC126901).

Can GPs use it for dictation or patient recordings?

That is what it is built for: dictation and sensitive recordings transcribed on hardware we own, with no audio or transcript kept on our servers afterwards. See secure AI for GPs for the clinical-workflow side.

What does Hush Echo cost?

Echo is included in every Hush AI plan alongside chat, document review and the other tools. Plans start with a free two-week trial; see plans and pricing.

Related

Secure AI for GPs · Private AI for solicitors and law firms · Security architecture · Plans and pricing