Get API accessBook a demo
Developer Solutions

European speech AI, available in one line of code.

The Voxist API exposes ASR, translation, and voice synthesis via REST and gRPC. EU-hosted, EUR-billed, GDPR-native by architecture. No round-trip to a US hyperscaler.

View the docsStart for free
Integrated inOVHcloudScalewayVoxlive CCaaSELLIOT (Horizon Europe)
Quick exampleREST
POST /v1/asr/stream
Authorization: Bearer <api-key>
Content-Type: audio/webm

→ {
  "transcript": "Pouvez-vous confirmer le délai ?",
  "lang": "fr",
  "confidence": 0.97,
  "latency_ms": 186
}
REST + gRPCEU-hosted<200ms
The problem

Three problems other speech APIs don't solve

Problem 1

Your speech API was trained on English — and it shows on French

Most major ASR APIs have 8–15% WER on general French, and worse on accented French or technical terminology. The Voxist API reaches 4.2% WER on French — best-in-class for European languages.

Problem 2

Your translation API pivots through English — and that flattens the nuance

French → English → German loses 30% of the formal nuance in the English pivot. The Voxist API translates directly between European pairs — COMET #2 globally on French-centric EU pairs, no pivot.

Problem 3

Your DPO is asking where your users' audio goes

Applications processing voice from EU residents must answer strict GDPR requirements on data location. The Voxist API is hosted on OVHcloud and Scaleway, in the EU, with zero transit to the US.

What you can build

Everything your application can do with voice

Speech recognition (ASR)

Streaming and batch. 40+ languages. Automatic language detection. 4.2% WER on French. Technical vocabulary available.

Translation (MT)

Streaming and on-demand translation. 1600+ language pairs. COMET #2 globally on French-centric EU pairs. Code-switching detected automatically.

Neural voice synthesis (TTS)

Natural neural voices in 40+ languages. Real-time streaming. Prosody control. Voice cloning on request (enterprise contract).

Voice-to-voice pipeline (VTT)

ASR + MT + TTS as a single endpoint. Perceived latency under one second. Ideal for real-time translation applications.

Multi-language SDKs

Python, Node.js, Go, Rust, Java, and .NET. WebSocket clients for streaming. OpenAPI spec for your own client generation.

On-premise option

For applications that cannot send audio to the cloud. Models deploy in your infrastructure. Same API, same SDK.

Compliance & trust
GDPR-nativeEU AI Act readyEU-hosted (OVHcloud · Scaleway)SOC 2 Type II (in progress)ISO 27001 (in progress)No audio for model trainingOn-premise option
FAQ

Questions, answered

Where is the Voxist API hosted?
The Voxist API is hosted on OVHcloud and Scaleway, exclusively in EU datacenters. No audio or metadata transits through the US or any other third country.
Is my audio used for model training?
No. No audio submitted via the API is used to train or improve Voxist models. Data is processed in-flight and not persisted unless you explicitly configure retention (for your own transcripts).
What audio formats does the API support?
WebM/Opus, WAV PCM 16-bit, MP3, FLAC, OGG. For real-time streaming, WebM/Opus and WAV PCM are recommended. Minimum sample rate is 16kHz; 48kHz recommended for translation.
Does the API have rate limits?
The free plan is limited to 5 req/s and 5h audio/month. The pay-as-you-go plan supports 50 req/s by default, increasable on request. The enterprise plan has no rate limit.
Can I use the API in production before upgrading to a paid plan?
Yes. The free plan includes a 99.5% SLA and is suitable for moderate production load. Upgrade to pay-as-you-go or enterprise when your needs exceed 5h/month or 5 req/s.

Build on speech AI that will stay where your data needs to be.

Start building free

Free API key · EU-hosted · no audio used for model training