# shiningstar1128/monkey_01

QWEN3Vox voice TTS for the [Vocence](https://github.com/vocence-78) subnet.

## Hugging Face repo id

`model_name` in `vocence_config.yaml`, on-chain commit, and chute wrapper `VOCENCE_REPO` must all be:

**`shiningstar1128/monkey_01`**

## Validator `/speak` contract

Validators POST JSON (natural-language prompts after source extraction):

```json
{
  "text": "<transcription to synthesize>",
  "instruction": "An adult female with an American accent, speaking at a normal pace in a mid-range pitch, sounding neutral and formal throughout"
}
```

- `text` — passed verbatim as the synthesis script (length cap only).
- `instruction` — passed verbatim to the processor `system_prompt` (length cap only).

**How this engine uses them:**

- `instruction` → processor `system_prompt` (no trait parsing or paraphrasing).
- `text` → script body (plain transcript or existing `Speaker N:` lines).
- Default aux audio shard for acoustic prefill only.

## Repo layout

| File | Purpose |
|------|---------|
| `miner.py` | QWEN3Vox engine (`Miner` class) |
| `vocence_config.yaml` | `model_name` + runtime/generation limits |
| `chute_config.yml` | Chutes image, GPU (64 GB VRAM), pip deps |
| `model-0000*-of-00003.safetensors` | Sharded weights (≥ 50 MiB total) |
| `aux_lm_residual_projection.safetensors` | Aux conditioning shard |
| `tokenizer.json`, `tokenizer_config.json`, `vocab.json`, `merges.txt` | Tokenizer |

## Deploy

1. Upload this folder to `shiningstar1128/monkey_01` and pin a **40-char commit SHA**.
2. Render/deploy chute with `VOCENCE_REPO=shiningstar1128/monkey_01`.
3. Commit: `uv run vocence miner commit --model-name shiningstar1128/monkey_01 --model-revision <sha> --chute-id <uuid>`.

See [miner_sample/MINER_GUIDE.md](../../../miner_sample/MINER_GUIDE.md) for anti-cheat rules.