You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

sl-antelope

OmniVoice (k2-fsa) LoRA adapter fine-tuned on Tajik (khazina_tg_24k_REAL, 661.4h/217,532 clips), official LoRA recipe (r16/a32, 5000 steps, lr1e-4). n=60 frozen-set WER: mean 7.19%/median 5.56% vs zero-shot 7.85%/6.25% and Chatterbox cb_step8000 10.73%/8.33%. LICENSE: base weights CC-BY-NC (k2-fsa/OmniVoice) -- this adapter inherits NC. Research/paper use only, NEVER for the Sado product. Run record: Tohirju/omnivoice-tts-eval-staging (run 2026-09-21_omnivoice-tg-lora_tts_r1).

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support