File size: 770 Bytes
2094eb1
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
# voicedesign_t0

PromptTTS model based on **Qwen3-TTS-12Hz-1.7B-Base**, with small updates from fine-tuning on additional training data.

## Contents
- `miner.py` — Vocence PromptTTS engine (`class Miner`: `__init__`, `warmup`, `generate_wav`)
- `chute_config.yml` — Chutes build config (image, node selector, chute settings)
- `vocence_config.yaml` — runtime options (sample_rate, adapter, limits)
- `model.safetensors` — fine-tuned model weights
- `speech_tokenizer/` — RVQ speech tokenizer
- `tokenizer_config.json`, `vocab.json`, `merges.txt` — text tokenizer assets

## Training notes
Small incremental update over the base model, fine-tuned on a modest instruction-annotated dataset. Intended for Vocence subnet (Bittensor SN78) PromptTTS evaluation.