COCKTAIL
A tool that generates new training data with varied voices,
phrasing, and overlapping channels to simulate noise
A tool that generates new training data with varied voices, phrasing, and overlapping channels to simulate noise
Communications with Operational Context and Knowledge for Target Audio Identification Learning (COCKTAIL)
Charles River Analytics developed COCKTAIL to automatically generate realistic speech and chat data for training simulations. Using advanced natural language processing and text-to-speech techniques, COCKTAIL enables instructors to create dynamic, high-variability communication scenarios that better reflect real-world operational environments.
Originally applied to air traffic control, COCKTAIL is now focused on supporting maritime ISR training, where operators must identify critical communications amid overlapping channels and background noise. The system is designed to extend across domains and can be adapted to integrate with a wide range of training simulators.
Developed with support from the Naval Air Systems Command Naval Air Warfare Center Training Systems Division (NAWCTSD), COCKTAIL addresses a persistent challenge in training systems: static or reusable audio that trainees learn to recognize, reducing training effectiveness.
“When we improve how communication is generated for training, we create more realistic environments, and more prepared operators.”
Dr. Michael Giancola
AI Scientist and Principal Investigator on COCKTAIL
Generating Realistic, Variable Training Data
COCKTAIL enables instructors to generate high-quality speech and chat data tailored to specific training needs. Instructors define key parameters such as:- Entities participating in the conversation
- Topic and context of communication
- Voice characteristics and background noise
Audio is generated using state-of-the-art text-to-speech models, producing natural, human-like speech suitable for immersive training environments. A timeline-based interface allows instructors to:
- Arrange communications in sequence
- Control the degree of overlap between channels
- Simulate multi-speaker, noisy operational conditions
Each scenario is saved as a reusable “recipe,” enabling instructors to quickly regenerate new variations with different voices, phrasing, and call signs—preventing trainee familiarity and maintaining training effectiveness.
Flexible Integration Across Training Systems
COCKTAIL features a modular architecture designed to integrate with a wide range of training simulators and applications. The platform can be adapted to:- Support multiple training environments
- Interface with different simulation systems
- Extend across domains such as maritime, aviation, and other operational settings
COCKTAIL can be deployed as a standalone tool or integrated as a backend service for generating speech and chat data within larger training ecosystems.
Transition and Partnership Opportunities
COCKTAIL is positioned for transition into operational and commercial training environments.We are seeking partners to support continued development and deployment, including:- Licensing COCKTAIL as a standalone solution
- Integrating COCKTAIL into existing training systems
- Expanding capabilities for new domains and use cases
COCKTAIL provides a scalable approach to generating realistic communication environments, supporting more effective and adaptable training.
COCKTAIL supports target audio identification training across defense and other operational domains, with flexibility to extend into additional applications. It also has the potential to enhance Charles River’s KWYN® suite of training solutions and other commercial systems that require scalable, realistic speech and chat generation for training.
To learn more, watch Mike’s Navy STP Tech Talk to see how COCKTAIL generates realistic, variable communications for training simulations in complex environments.
Contact us to learn more about COCKTAIL and our other adaptive training and artificial intelligence capabilities.
This material is based upon work supported by the Naval Air Systems Command (NAVAIR) under Contract No. N68335-23-C-0300. Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of NAVAIR.
On the job, the KWYN® AI platform augments skills through intelligent guidance and coaching.