SCRIBAL: AI Infrastructure for Real-Time Transcription, Translation and Accessibility
Advantages
- Precision: Error rate reduced to less than half compared to standard base systems.
- Linguistic Freedom: Facilitates multi-cultural collaboration by allowing speakers to use their native language.
- Data Sovereignty: Self-managed option (On-Premises) provides maximum control over privacy and data security.
- Efficiency: Minimizes problem-solving time and automates the creation of documentation and summaries.
- Adaptability: Specifically optimized for local dialects and technical/university terminology.
Goal
The group is looking for license partners, but other collaborations may be considered.
Intellectual Property
This Project is protected by intellectual property registration.
Reference
UBTT0553
Contact
Isabel Durán, PhD
Email: iduran@fbg.ub.edu
SCRIBAL: AI Infrastructure for Real-Time Transcription, Translation and Accessibility
Executive summary
A research group from the University of Barcelona, specialized in artificial intelligence and language technologies, has developed an automatic Speech-to-Text (STT) transcriptionist designed for high-demand technical environments. Scribal allows fast and accurate transcription and simultaneous translation of classes, conferences, and professional meetings. It addresses the needs of both the educational sector (accessibility) and the consultancy market (efficiency and linguistic freedom).
Introduction
Currently, most educational institutions and professional organizations face communication barriers due to language limitations and the high costs of manual transcription. This deficit restricts accessibility for students with hearing impairments and limits productivity in multi-cultural professional teams. In the consultancy and corporate world, knowledge shared during meetings is often «ephemeral» and difficult to index. Furthermore, speakers often limit their nuance and fluency when forced to speak in a non-native language. Scribal provides a solution that transforms these organizations into data-driven and inclusive environments.
Description
Scribal is an advanced AI infrastructure for real-time transcription, translation, and accessibility. The technology is based on the OpenAI Whisper system, optimized with neural networks and trained with specific databases to adapt to different dialects (specifically Catalan varieties) and specialized terminology.
- Accessibility: Real-time subtitles for students with hearing impairments or non-native speakers.
- Study tools: Synchronization of text with audio and exportable transcripts for study materials.
- Linguistic Freedom: Allows instructors, mentors, and team members to speak in their native language while participants receive immediate translation (Spanish, English, Chinese, etc.), eliminating language barriers.
- Automated Documentation: Automates the generation of summaries and meeting notes, offering operational savings.
- Content Generation: Automatically generates assets for SEO, blogs, and social media from training content.
The system offers flexible deployment, including SaaS Cloud or On-Premises solutions, ensuring total sovereignty and high security for sensitive data.
Current stage of development
The project is in an advanced stage of technological maturity (TRL 9). Ready to market, with a roadmap for implementation with initial clients and expansion in 2026.