Blog

9 July 2026

Introducing Crosstalk 1 and Crosstalk 1 mini

A new family of text-to-speech models optimized for mobile hardware.

2 July 2026

What do LLMs think they sound like?

A little experiment with Qwen3-TTS Voice Design and LLM self-description.

29 June 2026

Generating images with a 2025 Android

Running PrismML's Bonsai Image model on a 2025 Android phone's Hexagon NPU.

26 June 2026

Denoising voice recordings on-device

Cleaning audio input for voice cloning with a denoiser ported to run fast on mobile.

23 June 2026

Generating images with a 2020 iPhone

Running PrismML's Bonsai Image model on a six-year-old iPhone, with mlx-swift.