Qwen3-TTS but in OpenVINO, from scratch

Hello everyone, I finally got around to preparing my implementation of Qwen3-TTS in OpenVINO format as a codebase. This work was done in early 2026, merged to OpenArc in March and I kept forgetting about releasing the code. Here we are. https://github.com/SearchSavior/Qwen3-TTS-OpenVINO One guy from our discord speaks russian and I wanted to voice clone elmo on my A770,so I decided to from scratch Qwen3-TTS in pytorch, ignoring transformers (except for AutoTokenizer, my beloved) to really get