Reson8, the European speech technology company building hyper-customised automatic speech recognition (ASR), today announced it has raised a €5 million pre-seed round led by Balderton Capital, with participation from NP Hard, as it prepares to bring a new generation of speech infrastructure to market.
While English speech recognition is often considered a solved problem, performance drops sharply across European languages, dialects and industry-specific jargon, forcing teams to rely on manual correction or generic models that guess rather than listen. Reson8 is tackling the gap between expectation and performance with a new approach to speech AI; building acoustically precise models that can be customised in real-time using live context such as documents, websites and calendars. Instead of retraining large models or layering language models on top, Reson8 applies small, pluggable adapters that tune speech recognition to the specific context of each conversation, improving accuracy without adding latency or guesswork.
Founded by three deeply technical builders – Thomas Kluiters, Raoul Ritter and Jarno Verhagen – the Reson8 team came together around a shared frustration with generic, one-size-fits-all speech AI. Thomas and Jarno previously worked at high-frequency trading firm IMC, where performance and precision were non-negotiable, before taking those instincts into speech technology. Thomas also co-founded healthcare speech company Juvoly, acquired earlier this year, where Jarno was also CTO. Raoul brings an AI and large-scale systems background from Adyen.
Solving “your language, your jargon” where generic ASR fails
Reson8 is built on the belief that speech recognition should adapt to people, rather than forcing users to adapt to software. Instead of optimising for broad, average performance, the company prioritises acoustic precision and control, allowing its models to reflect how language is actually spoken in real settings. The result is speech recognition that handles specialist terminology, names and fast-moving dialogue with far less friction, and can be tailored quickly to different professions, organisations or even individual conversations.
The approach works wherever language gets specific:
- Healthcare: Handles drug names, clinical shorthand, and the precise terminology of every specialty, the words generic models consistently mishear, misread, or miss entirely.
- Legal: Adapts to firm-specific language, contractual terms, and jurisdiction-specific vocabulary where a single mistranscription can change the meaning of an agreement.
- Customer Support: Adapts instantly to product names, technical terms, and brand-specific language that define every support interaction – so the right words are always recognised correctly.
Built and hosted in Europe by design
Unlike many speech providers that depend on US hyperscalers, Reson8 is building and operating its own European infrastructure. This approach gives customers full-stack ownership, clear data residency, and an advantage as regulation and procurement requirements increasingly favour European-hosted AI.
At launch, Reson8 will include support for more than 20 European languages, addressing a persistent quality gap in non-English speech recognition.
Funding and growth plans
The €5 million pre-seed round will be used to expand Reson8’s European hardware footprint, continue development of its inference stack and foundational speech models, and grow the team selectively, prioritising talent density over headcount.
Raoul Ritter, co-founder of Reson8, said: “Generic speech models break the moment you leave English, and customers increasingly want control over where their data lives. We feel strongly that Europe doesn’t need to be a fast follower in foundational AI. Instead, we can build the speech layer Europe can rely on: deeply customisable, engineered for accuracy, and run on infrastructure we own.”
James Wise, Partner at Balderton Capital, said: “Speech is quickly becoming a default interface across software, from productivity tools to healthcare and public services. The Reson8 team stands out for their technical depth, the product’s focus on accuracy in real-world situations, and its Europe-first approach to infrastructure and data. We’re excited to back the team as they build a category-defining global speech company.”
About Reson8
Founded in Amsterdam, Reson8 is building hyper-customised speech recognition that adapts to each conversation in seconds. Its platform enables companies to deploy specialised speech models with greater accuracy across European languages and industries.
About Balderton Capital
Balderton Capital is a multistage venture firm with more than 25 years of experience supporting Europe’s best founders from Seed to IPO. We have both early and growth funds and invest across the technology sector, with a proven track record backing AI, fintech, B2B SaaS, digital health, mobility, gaming and marketplace companies. Previous investments include Darktrace (LON: DARK), Depop, Dream Games, Flywire (NASDAQ: FLYW), Kobalt, MySQL, Nutmeg, Peakon, Recorded Future, Talend (NASDAQ: TLND) and THG (LON: THG). Balderton’s current portfolio includes: Aircall, Contentful, Cleo, Fuse, GoCardless (Mollie), Lendable, Matillion, Photoroom, Primer, Proxima Fusion, Quantum Systems, Revolut, The Exploration Company, Tibber, Wayve, Writer and ZOE.
