Deepslate Secures €7.7M Seed Funding to Build European Speech-to-Speech AI Models
Deepslate, a Brandenburg-based AI model lab headquartered in Templin, has secured €7.7 million in Seed funding to advance speech-to-speech AI models built and hosted entirely in Europe.
Deepslate, a German AI model lab, has raised €7.7 million in seed funding to develop speech-to-speech AI models hosted entirely in Europe.
What happened
Deepslate, a Brandenburg-based AI model lab, has raised €7.7 million in seed funding to advance speech-to-speech AI models built and hosted entirely in Europe. The company's end-to-end models process spoken language directly, reducing latency and preserving tone, emphasis, and dialect. Their system includes an in-house speech encoder, a reasoning core using an open-weights language model, and a speech decoder. The model achieved a 440-millisecond response time, the fastest for speech-to-speech as of September 2026, and posted the best error rate for European languages in the CoVoST2 comparison. The funding will support expanded model training, a German-language data program, and growth in sales, marketing, and European production infrastructure.
Why it matters
The development of speech-to-speech AI models with low latency and high accuracy is crucial for applications in customer service, insurance, and other industries requiring real-time, natural language processing. European data sovereignty ensures compliance with local regulations and builds trust among users.
- Deepslate's speech-to-speech models offer faster response times and better accuracy for European languages.
- European data sovereignty is a key feature, ensuring compliance and trust.
- The funding will support further development and expansion of the technology.
The bigger picture
The advancement of speech-to-speech AI models with European data sovereignty can enhance real-time communication technologies, improve customer service experiences, and ensure compliance with data regulations, making them more attractive for European businesses and users.
Deepslate has secured significant seed funding to develop speech-to-speech AI models in Europe, focusing on low latency, high accuracy, and data sovereignty.
Technical details
- End-to-end speech-to-speech models that process spoken language directly without text conversion.
- In-house speech encoder, open-weights language model post-trained for European languages, and speech decoder.
- 440-millisecond response time in an independent benchmark.
- CoVoST2 comparison for error rate in European languages.
FAQ
What is Deepslate's main focus?
Deepslate focuses on developing end-to-end speech-to-speech AI models that process spoken language directly, reducing latency and preserving tone, emphasis, and dialect.
What are the key features of Deepslate's AI models?
The models use an in-house speech encoder, a reasoning core with an open-weights language model, and a speech decoder. They achieve a 440-millisecond response time and have the best error rate for European languages in the CoVoST2 comparison.
Why is European data sovereignty important for Deepslate?
European data sovereignty ensures that all hosting is within the EU, meeting customer requirements for compliance and data security.
This summary and analysis were generated with AI from the original article and may contain errors. Read the original for the full story.
Read the full story
Originally published by The AI Insider on Oct 7, 2026.
Comments
Sign in to join the discussion.