COMPRISE · Project

Privacy-First Multilingual Voice Technology That Cuts Development Costs for Businesses

digitalTestedTRL 6

compriseh2020.eu

Imagine you want to add a voice assistant to your app — like Siri or Alexa, but in Latvian or Catalan. Today, you'd need mountains of recorded voice data stored in the cloud, which is expensive, raises privacy red flags, and only works well for major languages. COMPRISE built a toolkit that lets companies create voice-enabled services without sending user voices to the cloud, supports smaller languages out of the box, and uses smart automation to slash the cost of training these systems. Think of it as a privacy-safe, budget-friendly shortcut to building your own multilingual voice interface.

By the numbers

EUR 3,201,016

EU funding for privacy-first voice technology R&D

consortium partners across 4 countries

SMEs involved in development

sectors demonstrated (smart consumer apps, e-commerce, e-health)

total project deliverables produced

50%

industry ratio in consortium

The business problem

What needed solving

Building voice-enabled services today means sending user voice data to cloud servers controlled by big tech companies, creating serious privacy risks and GDPR headaches. It also costs a fortune to support smaller European languages because you need massive labelled datasets for each one. European businesses are stuck choosing between expensive, privacy-risky solutions from US tech giants or going without voice features entirely.

The solution

What was built

The team built a complete SDK (software development kit) with online documentation and a cloud-based resource platform for creating multilingual voice services. They delivered a final platform demonstrator with trained models, validated against GDPR requirements, and demonstrated across 3 commercial sectors. A total of 23 deliverables were produced including 4 demo-level outputs.

Audience

Who needs this

E-commerce platforms adding multilingual voice search for European marketsHealth tech companies building GDPR-compliant voice interfaces for patient interactionCustomer service software vendors expanding language coverage affordablySmart home and IoT device makers needing privacy-safe voice controlGovernment digital service teams building multilingual citizen-facing voice portals

Business applications

Who can put this to work

E-commerce

any

Target: Online retailers and marketplace platforms serving multilingual European customers

If you are an e-commerce platform dealing with customers who speak different languages and expect voice-based product search — this project developed an SDK and cloud platform that lets you build multilingual voice shopping features without storing customer voice data, covering languages that big tech voice assistants ignore. The toolkit was demonstrated specifically for e-commerce use cases across the 8-partner consortium.

Healthcare

mid-size

Target: Telemedicine providers and health app developers

If you are a health technology company dealing with GDPR compliance headaches when handling patient voice data — this project developed a privacy-by-design voice platform that transforms speech data so it cannot be traced back to individuals. The final platform demonstrator was populated with trained models and validated against data protection and GDPR requirements, making it directly relevant for e-health voice services.

Customer Service

SME

Target: Call center software vendors and chatbot providers

If you are a customer service technology provider dealing with the high cost of supporting voice interactions in multiple European languages — this project developed automatic labelling and integrated translation tools that reduce the data and manual effort needed to launch new language support. The SDK prototype integrates research on personalised learning and weakly supervised deep learning to cut voice technology costs.

Frequently asked

Quick answers

How much would it cost to implement this voice technology?

The project produced an open SDK and cloud-based resource platform, suggesting the tools may be available at low or no licensing cost. The full research effort was backed by EUR 3,201,016 in EU funding across 8 partners, so the underlying R&D investment is substantial. Specific per-unit or subscription pricing is not detailed in the available project data.

Can this scale to handle enterprise-level voice traffic?

The final platform demonstrator was deployed in a working environment and populated with trained models, indicating it moved beyond lab testing. However, the project focused on demonstrating feasibility in 3 sectors (smart consumer apps, e-commerce, e-health) rather than proving high-volume production throughput. Based on available project data, additional engineering would likely be needed for enterprise-scale deployment.

What is the IP and licensing situation?

COMPRISE was a Research and Innovation Action (RIA) funded under Horizon 2020, which typically means results are owned by the consortium partners. The SDK prototype and platform were developed by 8 partners including 4 industry players and 4 SMEs. Businesses interested in licensing should contact the coordinator (INRIA, France) to discuss access terms.

How does this handle GDPR and privacy regulations?

Privacy-by-design is the core principle of the entire project. The team developed privacy-driven data transformations that prevent voice recordings from being traced back to individual users. The final deliverable explicitly included updated data protection and GDPR requirements validation, making this one of the few voice technology toolkits built from the ground up with European privacy law in mind.

What languages are supported?

The project targeted cost-effective multilingual support, specifically aiming to serve under-resourced languages and accented speakers that major voice platforms ignore. The consortium spanned 4 countries (Germany, Spain, France, Latvia), suggesting at least those languages were covered. The integrated translation component was designed to let users access services in their own language regardless of what language the content was originally in.

How long would integration take?

The project delivered a documented SDK (with Swagger online documentation) designed to be easy to use for developers. The initial prototype was ready mid-project, with the final version integrating all research results. Based on available project data, a development team familiar with voice technology could begin integration using the SDK, though timeline depends on the specific use case complexity.

Consortium

Who built it

The COMPRISE consortium is well-balanced for technology transfer, with 8 partners split evenly between industry (4) and academia/research (4), and a 50% industry ratio — one of the better ratios you'll see in EU research projects. The 4 SMEs in the consortium signal that the technology was developed with commercial viability in mind from day one, not just as an academic exercise. The coordinator is INRIA (France), one of Europe's top computer science research institutes, which adds credibility. The consortium spans 4 countries (Germany, Spain, France, Latvia), giving it geographic diversity across both large and smaller EU markets — particularly relevant for a multilingual technology. The mix of research depth and commercial intent makes this consortium a solid candidate for technology licensing or partnership discussions.

INSTITUT NATIONAL DE RECHERCHE EN INFORMATIQUE ET AUTOMATIQUECoordinator · FR
ROOTER ANALYSIS SLparticipant · ES
UNIVERSITE DE LORRAINEthirdparty · FR
NETFECTIVE TECHNOLOGY SAparticipant · FR
TILDE SIAparticipant · LV
ASCORA GMBHparticipant · DE
UNIVERSITE DE LILLEthirdparty · FR
UNIVERSITAT DES SAARLANDESparticipant · DE

How to reach the team

The coordinator is INRIA (Institut National de Recherche en Informatique et Automatique) based in France. SciTransfer can facilitate a direct introduction to the project team.

Order a report on this consortium See INSTITUT NATIONAL DE RECHERCHE EN INFORMATIQUE ET AUTOMATIQUE profile →

Next steps

Talk to the team behind this work.

Want to explore how COMPRISE's privacy-first voice SDK could work for your product? SciTransfer can arrange a technical briefing with the development team and help you evaluate licensing options. Contact us for a free one-page solution brief.

Order a report on this topic More in Digital & ICT