Imagine having a digital assistant in your video calls that doesn't just transcribe words, but acts like a professional diplomat. It breaks down language barriers in real-time and reads body language to make sure everyone is on the same page. It's like having a universal translator and a meeting secretary rolled into one, ensuring no one is left out because of a language gap or a bad connection.
Virtual meetings are hindered by language barriers, poor accessibility, and a lack of trust in AI transparency. These frictions lead to isolation and inefficiency in global business collaboration.
The project built two cornerstone AI models, SpeechLMM and Mumospee, and a dataset with over 500,000 hours of audio. It also developed virtual assistants named Agentar and Butler for real-time meeting support.
If you are a law firm dealing with clients across different EU countries and language barriers — this project developed real-time speech-to-speech translation and summarization that ensures legal nuances are captured accurately. This reduces the need for expensive human interpreters in every preliminary meeting.
If you are an EdTech provider dealing with fragmented accessibility features for diverse students — this project developed multimodal AI agents like Agentar and Butler that support inclusive interaction. This allows students from different linguistic backgrounds to engage in the same virtual classroom seamlessly.
If you are a global company dealing with 'techno-clutter' and inefficient virtual meetings across time zones — this project developed intelligent assistants for real-time summarization and mediation. This saves executive time by distilling long multilingual discussions into actionable insights.
Based on available project data, no specific pricing or cost structure has been disclosed as the project is currently in the research and development phase.
The project uses foundation models and a massive dataset of over half a million hours of speech, suggesting a design intended for large-scale industrial application across Europe.
Based on available project data, the project emphasizes open data and open models, though specific commercial licensing terms are not yet detailed.
Yes, the project is specifically designed to comply with the EU AI Act and the Digital Decade strategy to ensure trustworthiness and accountability.
The project period runs from 2024-01-01 to 2027-12-31, indicating that full integration and deployment targets are aligned with this timeline.
The consortium is heavily industry-weighted with a 62% ratio, comprising 5 industrial partners (including 2 SMEs) and 3 academic/research entities across 5 countries. This strong commercial presence, led by an Italian SME (Translated SRL), suggests a high priority for market translation and practical application rather than purely theoretical research.
Contact Translated SRL in Italy for partnership inquiries
Contact us to explore licensing opportunities for the SpeechLMM and Mumospee models.
