The Challenge

Brilliant Labs set out to deliver a seamless, real-time audio experience from smartphone to smart glasses - requiring high-quality text-to-speech that could run efficiently across constrained devices.

However, deploying advanced AI voice models in this environment posed key challenges:

• Model size constraints limited on-device feasibility

• Latency requirements demanded near-instantaneous speech generation

• Cross-device performance needed to remain consistent from phone to wearable hardware

Our Solution

Neuphonic partnered closely with Brilliant Labs to adapt our AI TTS technology for this unique hardware setup.

We reduced the model size by 10x using distillation and pruning, then quantized it for efficient deployment. We also worked with TheStageAI to optimize the inference runtime and integrated the model into a phone-to-glasses pipeline.

The Results

The result was a high-performance, real-time voice system that met the demands of wearable computing.

The system delivers real-time, ultra-low-latency speech with seamless performance from phone to glasses, all while running on a dramatically reduced compute footprint with no loss in voice quality.