
Deepslate
European speech-to-speech voice AI model for real-time voice agents, hosted in the EU.

About Deepslate
Deepslate
Deepslate is a speech-to-speech voice AI model from Deepslate GmbH, a German company founded by Paskal Paesler and Jan Brachthäuser. Its Opal model takes audio in and returns audio out in 27 languages, replacing the usual STT, LLM and TTS chain, and runs on EU hosting or on premises. The company raised a €7.7M seed round led by 42CAP in October 2026.
Key Features
- Native speech-to-speech: One realtime model listens, reasons and responds, with 440 ms time to first audio per Artificial Analysis.
- 27 languages: Multilingual conversations with server-side VAD and native interruption handling.
- Developer integrations: Official LiveKit and Pipecat plugins, Python and Node SDKs, plus WebSocket, WebRTC and SIP.
- Voices and tools: Hosted voices and voice cloning, function tools and webhooks for live data lookups, and call logs.
- EU data residency: Processing and storage in the EU, ISO 27001 certification stated by the company, and self-hosting for enterprise customers.
Use Cases
- Insurance hotlines: Answer policy questions on the phone, starting from the ready-made insurance FAQ template.
- Contact centers: Handle high-volume inbound and outbound calls over SIP with existing phone numbers.
- Voice agent builders: Swap a cascaded STT, LLM and TTS pipeline for a single realtime model inside LiveKit or Pipecat.
Getting Started
Sign up for a free account with €15 in credits and call your first agent in the browser: https://deepslate.eu

