
AuK: An Open-Source Foundational Model for Speech Generation and Editing
AuK: An Open-Source Foundational Model for Speech Generation and Editing
Verify evals on Papers with Code
[Question/Feature] Non-EN/ZH languages: measured zero-shot Polish result (WER ~80 % vs 23 % for English in the same voice) and fine-tune recipe for adding a language
[Feature] Native MLX backend for Apple Silicon
[Feature] Variant on a permissively licensed encoder — the required Qwen2.5-Omni-3B is research-only, which makes the MIT release non-commercial in practice
[Feature] Multi-GPU placement and two-node split deployment