---
license: apache-2.0
language:
- en
base_model:
- Gryphe/Gemma-4-12B-StyleTune
tags:
- gemma4
- styletune
- voice
- gguf
- roleplay
- creative-writing
- style-transfer
- qwen
- llama
library_name: voice
pipeline_tag: text-generation
---

<p align="center"><strong>The StyleTune voice for Gemma 4 12B — as a download.</strong></p>

# Gemma-4-12B-StyleTune-Voice

**56% fewer clichés. 16.8% shared vocabulary. One tensor. ~2.0 GB.**

This is not a model. It is the *voice* of [Gryphe/Gemma-4-12B-StyleTune](https://huggingface.co/Gryphe/Gemma-4-12B-StyleTune), extracted into a single portable tensor. Pair it with the [Voice tool](https://huggingface.co/Wiself/voice) and any **Gemma 4 12B** GGUF you already have becomes StyleTune-voiced — reasoning, knowledge, and instruction following untouched.

---

## Why this exists

A full StyleTune finetune is 24 GB. The voice is **2.0 GB** — and it carries the entire style. The style lives in one tensor: the `lm_head` output projection, the last stop before text appears. Gryphe trained only that tensor; we lift only that tensor.

Download the voice, not the model.

---

## Two steps

```bash
# 1. Get the voice tool (one-time): https://huggingface.co/Wiself/voice
python3 voice.py path   # optional, puts `voice` on your PATH

# 2. Cast it onto any Gemma 4 12B GGUF you already have
voice cast ./gemma-4-12b-it-Q4_K_M.gguf voice.safetensors --out ./voiced/gemma-4-12b-styletune.gguf
```
Append `--speak` to force the voice through `output.weight` (invented if the target lacks it and geometry survives; GGUF only).

Run it:

```bash
llama serve -m ./voiced/gemma-4-12b-styletune.gguf --jinja
```

That's it. One file out, nothing extra at runtime. No adapter, no LoRA, no second model.

---

## The numbers

From Gryphe's own benchmarks ([source card](https://huggingface.co/Gryphe/Gemma-4-12B-StyleTune)) — 200 diverse roleplay prompts, greedy 0.0, versus the base instruct:

| Metric | Base | StyleTune |
|---|---|---|
| Clichés per 100 words | 1.050 | **0.463 (−56%)** |
| Shared trigram vocabulary | — | **16.8%** (an almost entirely different phrasing) |

Everything that makes StyleTune StyleTune is in this file. What does *not* change: reasoning, world knowledge, instruction following, multilingual capability — none of it lives in `lm_head`.

---

## What's inside

- `voice.safetensors` — the `lm_head.weight` tensor, **BF16**, shape `[262144, 3840]`, ~2.0 GB
- `voice.json` — metadata: source, dtype, shape, architecture

Bit-for-bit identical to the tensor that makes [Gryphe/Gemma-4-12B-StyleTune](https://huggingface.co/Gryphe/Gemma-4-12B-StyleTune) sound the way it does. If you have the original model, you don't need this file. If you have any *other* Gemma 4 12B — a quant, an abliterated variant, a different finetune — this voice drops in.

---

## Compatibility

| Target | Works? |
|---|---|
| Any Gemma 4 **12B** GGUF (any quant: Q4_K_M, Q5_K_M, Q8_0…) | ✅ |
| Abliterated / uncensored Gemma 4 12B GGUFs | ✅ (use the [delta variant](#loops--abliterated-targets) if it loops) |
| Gemma 4 **other sizes** (9B, 26B, 31B) | ❌ vocab/hidden mismatch — use the matching voice |
| Non-Gemma architectures | ❌ |

The cast checks shape before touching anything and tells you plainly if it won't fit.

<details>
<summary><strong>Loops on abliterated targets?</strong></summary>

On some abliterated targets a direct cast can loop. The fix is the delta variant — it carries only the *difference* from the base instruct, so MoE routing stays calm:

```bash
voice delta voice.safetensors          # needs the base: --base google/gemma-4-12b-it
voice cast ./model.gguf delta-voice.safetensors --out ./voiced/model.gguf
```

This is the exact path that fixed looping on the 26B QAT abliterated model — see the [Voice tool card](https://huggingface.co/Wiself/voice) for that story.

</details>

---

## Notes

- **Quality:** the voice is stored at the original BF16 precision. Casting to GGUF quantizes only the head to Q8_0 (≈¼ size, near-lossless); every other tensor is byte-copied from your model.
- **Casting to safetensors targets** works too — the head is replaced at the target's own dtype, no quantization.
- **Sampler tips from Gryphe:** temp 1.0, MinP 0.10, DRY sampler on. Gemma 4's native chat template applies automatically.
- Verify after download: `voice info voice.safetensors` → should read `lm_head.weight · [262144, 3840] · BF16`.

---

## Credits

[Gryphe](https://huggingface.co/Gryphe) trained the StyleTune and proved one tensor changes everything. The [Voice tool](https://huggingface.co/Wiself/voice) lifts and casts it. Gemma 4 is Google's model, under its own terms — this voice carries the source model's license lineage; check the [source card](https://huggingface.co/Gryphe/Gemma-4-12B-StyleTune) before sharing voiced models.