Solution · Medical transcription

Transcription at scale, without leaving the region.

Whisper-class transcription on the OpenAI audio shape, per-minute billed, pinned to Switzerland, the EU or the US with no cross-border failover.

Get started Read the docs

The problem

Why this is hard

Dictation and consult recordings are high-volume and unambiguously PHI. Transcription vendors often process audio wherever capacity is cheapest — which is exactly what a regulator asks about. phi-cloud gives you Whisper-class transcription that resolves to a same-jurisdiction host, bills by the minute, and returns the subtitle formats your tooling already expects.

How it works

The pipeline, end to end

1
Send the audio
POST /v1/audio/transcriptions
Multipart upload with model=whisper, X-Region and X-PHI:true. Choose json, text, srt, vtt or verbose_json depending on what you need downstream.
2
Stay in jurisdiction
Residency-separate
CH → Azure Speech (Switzerland); EU → Scaleway/Voxtral; US → Azure. There is no cross-border failover — a failed CH call is a CH error, not a quiet EU completion.
3
Summarise (optional)
POST /v1/chat/completions
Pipe the transcript into a resident chat model for a structured summary, action items or coding suggestions — same region, same PHI gate.
4
Pay per minute
x-phi-cost-micro
Billing is per minute of audio with the flat +10% gateway margin. Every response reports its exact cost.

Built on:Speech (STT)Summarisation (chat)

Why phi-cloud

What makes it compliant

Whisper-class accuracy

The whisper canonical maps to Voxtral, Whisper large-v3 and Azure Speech depending on region.

No cross-border failover

Speech is residency-separate by design; audio never completes in another jurisdiction.

Per-minute billing

Transparent, usage-based pricing with the cost reported on every call — no seats, no subscription.

Subtitle-ready output

srt, vtt and verbose_json drop straight into existing transcription tooling.

In code

A representative call

Transcribe a consult in the EU with timestamps and segments — pinned, PHI-gated, per-minute billed.

curl

curl https://phi-cloud.com/api/v1/audio/transcriptions \
  -H "Authorization: Bearer $PHI_API_KEY" \
  -H "X-Region: EU" -H "X-PHI: true" \
  -F file=@dictation.wav \
  -F model=whisper \
  -F response_format=verbose_json

response

x-phi-routed: scaleway-whisper-v3/EU/phi
x-phi-tier: phi
x-phi-cost-micro: 660   # ~3.3 min @ per-minute rate

Compliance posture

CH transcription → Azure Speech (Switzerland), HIPAA BAA + nFADP/GDPR.
EU transcription → Scaleway / Voxtral (EU), GDPR Art. 28 DPA.
US transcription → Azure Speech (US), HIPAA BAA.
Outside CH/EU/US, transcription returns no_route — it is intentionally limited to verified jurisdictions.

FAQ

Common questions

The call fails in-region. Speech is residency-separate with no cross-border failover — we will not complete a Swiss transcription on an EU host to paper over an outage. That is a deliberate compliance guarantee.

No. Speech-to-text is intentionally CH/EU/US only and returns no_route elsewhere, because those are the jurisdictions with verified PHI-eligible hosts.

Clinical documentation Speech capability Pricing

Ready when you are

Build medical transcription on a gateway that survives the audit.

Free to test. Prepaid credits when you go live. The residency and PHI posture is the same in production.

Get started See the compliance posture →

Free to test · Prepaid credits, no subscription · No data retained