expressive-speech

MoVE proposes a Mixture-of-LoRA-Experts architecture that preserves laughter and crying in speech-to-speech translation, achieving 76% NV retention with only 30 minutes of curated data.

Dramabox is an expressive text-to-speech model by Resemble AI that uses prompt-driven control for speaker identity, emotion, and delivery, with optional voice cloning via a 10-second reference. Built on the LTX-2.3 audio diffusion transformer, it is open-sourced on Hugging Face.