# super_10

QWEN3Vox TTS for the Vocence subnet. Weights and `miner.py` live in this repo; load via `from_pretrained("Gem1832/super_10")` per `vocence_config.yaml`.

Required at repo root: `miner.py`, `vocence_config.yaml`, `chute_config.yml`, `config.json`, sharded `.safetensors`, `aux_lm_residual_projection.safetensors`.

Optional: `voices/*.wav` for discrete conditioning (`VOCENCE_PREFER_DISCRETE_COEFF_DIR`).