Deepslate raises €7.7M to expand European speech-to-speech AI

Deepslate raises €7.7M to expand European speech-to-speech AI

Berlin-based Deepslate has raised €7.7 million in seed funding led by 42CAP, with participation from Alstin Capital, existing investor SIVentures and business angels. The company develops speech-to-speech models and hosts its technology entirely within the EU. It plans to expand training data for European languages, including German street names, personal names and dialects, improve latency and voice quality, grow sales and marketing, and scale production infrastructure in European data centers.

Deepslate competes at the model layer of voice AI, offering its technology through a self-service platform and API, with volume and self-hosting options. Its architecture combines an in-house speech encoder and decoder with an open-weight language model that it post-trains for specific languages; the reasoning model can be replaced without retraining the entire system. This makes language specialization and deployment control central to its competitive position. Deepslate reports a 440-millisecond response time in an Artificial Analysis benchmark and says it was the fastest speech-to-speech model measured as of September 2026. These are reported results, rather than evidence of performance across every customer workload.

For enterprise builders, the concrete evaluation is whether that architecture handles local names and dialects reliably under production call volumes while meeting hosting requirements. Deepslate says insurers, contact centers and platforms already use its technology in production, and it holds ISO 27001 certification. Investors should distinguish those adoption signals from commercial scale: the article provides no revenue, customer-count or retention figures. The funding supports model refinement and deployment capacity, but does not establish the durability of its differentiation.

#Deepslate #VoiceAI #SpeechToSpeech #EuropeanAI #SeedFunding