GEI§T– Generator for Emotionally Customizable Synthetic Voices

In media production today, editing news reports by hand for other media channels generates high costs. Shorter versions are usually produced for different channels, such as social media. This often also necessitates complete redubbing by the original speaker. The objective of the GEI§T project is to develop an automated method for creating different versions of an original report, whereby for the most part both the summarized content and the redubbing are automated with the help of synthesized voices, individual voice profiles as well as freely selectable tonality and emotionality. Automatic transcription into easy language should also be possible. This will make work processes easier and even more efficient.

Under the leadership of nachtblau GmbH, a company based in Hamburg that specializes in efficient process solutions in the field of media production, researchers are examining to what extent modern AI-based technologies can improve existing processes, facilitate a faster and more efficient use of content, and accelerate workflow.

In this context, the researchers at the Hochschule der Medien HdM from Stuttgart, one of the project partners, are focusing on media ethics and the legal issues resulting from new AI technologies. Researchers at Fraunhofer IDMT in Oldenburg are concentrating within the project on AI-based signal analysis and speech synthesis. In parallel, their colleagues in Ilmenau are developing technologies for the reliable labelling of synthetically generated material and the associated usage information. The aim here is to ensure the authenticity and legally compliant use of the artificially generated voices. At the end of the project, the technologies developed will be incorporated in a demonstrator.

If successful, the developments will be integrated into nachtblau’s existing products.

Further Information

 

Press Release / 23.5.2022

Seeing Speech

New algorithms from Fraunhofer IDMT form the basis for the »Dialogue Detection« in Steinberg Media Technologies’ latest version of its audio post-production software Nuendo. 

 

Presseinformation / 4.11.2021

Better understanding

Tonmeistertagung 2021: Fraunhofer IDMT presents solutions for analysing, evaluating and improving speech intelligibility.

 

SITA – Better sound, less noise!

SITA addresses the main factors for poor speech intelligibility along the whole distribution chain and aims to eliminate existing barriers for the widest possible variety of target groups, applications and hearing scenarios with the help of innovative software technologies.

 

Project SpeechTrust+

Reliably detect synthetic speech

The goal of SpeechTrust+ is to detect AI-based speech synthesis and voice distortion.

 

Project vera.ai

Verification Assisted by Artificial Intelligence

The project vera.ai aims to provide professional, trustworthy AI solutions against advanced disinformation techniques.

The project is funded by the Federal Ministry of Education and Research under funding code FKZ 01IS23014 A-C within the funding priority “Erforschung, Entwicklung und Nutzung von Methoden der Künstlichen Intelligenz in KMU (KI4KMU)".

Analysis and optimization of speech intelligibility

Our software solutions are able to measure, display and optimize speech intelligibility – automatically if needed.