Case Study: DeepZen achieves rapid, cost-saving emotion-rich audiobook production with Oracle Cloud Infrastructure

DeepZen Turns Text Into Emotion-rich Speech With Oracle Cloud

DeepZen is a UK startup that uses AI to convert text into emotion-rich, human-like speech to make audiobooks and other voice services affordable and widely available. Facing an audiobook market dominated by costly, time-consuming studio recordings and a need to train complex neural networks and NLP models, DeepZen needed a flexible, high-performance computing platform that could scale quickly to clone voices with realistic emotion and intonation.

DeepZen joined Oracle for Startups and built its platform on Oracle Cloud Infrastructure HPC, using bare-metal NVIDIA A100 GPUs and OCI auto-scaling to meet heavy training demands. The move delivered a 36% boost in model performance, cut training from seven to five days (saving roughly a month every three months), enabled a 10-hour audiobook to be produced in about an hour instead of 65 hours, and helped the company offer scalable SaaS/API services while licensing voices fairly to talent.

DeepZen

Kerem Sozugecer

Chief Technology Officer and Cofounder