Speech-to-text API hosted in France
Your voice data stays in France. Full stop.
01 — The situation
Your audio files cross the Atlantic
You need to turn audio into text. Meetings, medical consultations, hearings, interviews. You may already be using Google Speech-to-Text, Amazon Transcribe or Whisper through the OpenAI API.
Except that your audio files cross the Atlantic. They are processed on American servers, by companies subject to the Cloud Act. And in your sector — medical, legal, HR — that raises a real compliance problem.
- Your voice data passes through servers outside the EU
- The GDPR requires control over where personal data is processed
- Medical transcriptions contain sensitive health data
- Legal recordings are covered by professional secrecy
- Your DPOs and compliance officers are sounding the alarm
02 — Comparison
Big Tech or sovereign?
| Standard Big Tech APIs | GOODEV API (sovereign) |
|---|---|
| Data processed in the USA | Servers in France only |
| Subject to the Cloud Act | Outside Cloud Act jurisdiction |
| No real GDPR guarantee | GDPR compliant end to end |
| Changing terms of use | Data deleted after processing |
| Risk of data retention | No retention by default |
| Dependence on a foreign provider | A French contact, a French contract |
03 — Solution
The GOODEV solution
A speech-to-text API that runs entirely on French servers. The same performance as Big Tech, with sovereignty on top.
-
P.01
High accuracy
A speech-to-text engine optimised for French. 92 to 97% accuracy depending on audio quality. Technical vocabulary customisable by field (medical, legal, technical).
-
P.02
Real time or batch
Two modes: streaming transcription over WebSocket for real time (live subtitling, dictation), or audio file upload for batch processing (meetings, archives).
-
P.03
Speaker diarisation
Automatic identification of speakers. Each transcription segment is attributed to the right speaker. Essential for meetings with several participants and for minutes.
04 — API
A simple API
# Transcription d'un fichier audio
curl -X POST https://api.goodev.fr/v1/transcribe \
-H "Authorization: Bearer VOTRE_CLÉ" \
-F "file=@réunion.wav" \
-F "language=fr" \
-F "diarize=true"
# Réponse
{
"text": "Bonjour, on peut commencer...",
"segments": [
{"speaker": "A", "start": 0.0, "end": 3.2,
"text": "Bonjour, on peut commencer."}
],
"duration": 3542.8,
"language": "fr"
}
Three lines of code to transcribe a file. The API is documented, versioned, and compatible with every language over HTTP.
Accepted formats: WAV, MP3, FLAC, OGG, M4A, WebM. No size limit for files in batch mode. Optional webhook to be notified when the transcription is ready.
05 — Uses
Use cases
-
C.01
Medical
Transcription of consultations, operative reports, medical dictation. Medical vocabulary is built into the engine. Compatible with the confidentiality requirements of health data.
-
C.02
Legal
Transcription of hearings, depositions, client interviews. Professional secrecy respected — no data leaves French territory. Export in a structured format for integration into case files.
-
C.03
Meetings and committees
Automatic minutes with speaker identification. Minutes sent automatically to participants. No more manual note-taking during meetings.
-
C.04
Call centres and support
Transcription of incoming calls for quality control, satisfaction analysis and training. High volume, batch processing. Automatic anonymisation available.
06 — Compliance
GDPR and health data
Compliance is not a marketing argument. It is the reason this API exists.
- Physical servers in France, no multi-region cloud
- Data encrypted in transit (TLS 1.3) and at rest (AES-256)
- Audio files deleted after processing (configurable)
- No data retention to “improve the model”
- DPA (Data Processing Agreement) provided on request
- HDS-compatible hosting available for health data
Data sovereignty is not a luxury. It is a legal obligation for many sectors, and a strategic choice for the others. Transcribing your sensitive data on servers you do not control is a risk you do not need to take.
07 — Prices
Prices per hour of audio
Billed per hour of audio transcribed. The more you use, the cheaper it gets. No minimum subscription.
-
F.01
Starter
0,40 € /hour
Up to 50 hours a month. Ideal for testing or occasional use. All features included.
-
F.02
Professional
0,25 € /hour
From 50 to 500 hours a month. Speaker diarisation and custom vocabulary included. Responsive technical support.
-
F.03
Company
0,20 € /hour
Over 500 hours a month. Guaranteed SLA, dedicated infrastructure possible, HDS hosting on request.
Try the API for free
1 hour of transcription free, no bank card needed. Monthly billing, no hidden fees.
08 — Questions
Frequently asked questions
How accurate is the transcription?
92 to 97% depending on audio quality. The engine is optimised for French, including technical vocabulary. A custom dictionary can be added to improve accuracy in your field.
Is the infrastructure HDS compatible?
The standard infrastructure is hosted in France on secure servers. For projects subject to the HDS requirement, specific certified hosting can be set up. Contact me to discuss it.
Which audio formats are accepted?
All common formats: WAV, MP3, FLAC, OGG, M4A, WebM. The API also accepts real-time audio streams over WebSocket.
Can the different speakers be identified?
Yes. Speaker diarisation is available: the system automatically identifies the speakers and attributes each segment to the right one. Essential for meetings and interviews.
Switch to sovereign transcription.
Describe your transcription need. I will reply within 24 hours with a technical and pricing proposal.