Dream TeampresentNocturne
Nocturne Demo Video
Interpret your dreams...
Nocturne is an intriguing way to interpret, store and share dreams, with responsive and reactive visual and textual feedback provided by two custom trained AI models, offering two distinct interpretation styles.
We wanted to give users the opportunity to record and analyse their dreams in an engaging way, turning dream fragments into an ongoing journal that could be shared and revisited.
The Team
Mike Winnard
Tom Glencross
Marcus Gough
Zoltan Mozga
Seif Hok
Technologies
We used: Node, React, Chakra, TsParticles, Motion Framer, Firebase, FASTapi, Docker, Python, PyTorch, Transformers
React with Chakra UI provided accessible components while TsParticles and Motion Framer enabled us to quickly create engaging animations and visual effects. Firebase handled authentication and real-time data without complex server management. FASTapi deployment in a docker containerisation gave us rapid Python endpoints to interface with the ML models, while PyTorch and Transformers provided the foundation for custom LLM fine-tuning using a LoRA configuration.
Challenges Faced
Transitioning to Python for ML was a steep learning curve. Spiking new tech was limited by time constraints for the project’s launch.