ICTNLP

A model series capable of outputting both speech and text. Despite its name, it is based on the Qwen2.5 series.

We're starting to see text-image-audio open-source fine-tunes become popular. Here's a speech-to-text-speech model.