---
license: apache-2.0
base_model: nineninesix/gepard-1.0
library_name: gepard
pipeline_tag: text-to-speech
language:
- sl
tags:
- text-to-speech
- tts
- slovenian
- gepard
- nanocodec
- qwen3_5
- lora
---

# Gepard Slovenian TTS (80k)

A LoRA fine-tune of [**nineninesix/gepard-1.0**](https://huggingface.co/nineninesix/gepard-1.0) — a
~555M autoregressive, Qwen3.5-backbone TTS that generates NVIDIA **NanoCodec** audio tokens — adapted
to **Slovenian** on the ARTUR read-speech corpus.

Gepard is natively multilingual (EN/ES/PT/NL) and grapheme-based with a language-tag front-end, so
Slovenian is added by fine-tuning on `sl: <text>` prompts — **no phoneme front-end** (the Qwen3.5
tokenizer handles č/š/ž cleanly and the multilingual base learns Slovenian grapheme-to-sound internally;
using IPA would have discarded that multilingual transfer).

## Honest quality note

The model **speaks Slovenian with correct šumniki (č, š, ž)** and its prosody is fairly natural
(inherited from the multilingual base). **Overall quality is not yet top-tier** — this is a first scaled
run (80k utterances, 5 epochs) on a single read-speech corpus. Expect occasional mispronunciations,
variable pacing, and sensitivity to generation parameters. It is competitive on naturalness but a
phoneme-based system can still be crisper on pure pronunciation.

**Emotion:** there is some **content-driven emotional inflection** — the multilingual base was
pretrained on emotion-annotated speech (EmoNet-Voice), so emotionally-charged text and punctuation
(questions, exclamations, sad/angry phrasing) nudge the prosody, in clean Slovenian with no
foreign-accent leak. It is **mild, not cinematic**, and there is no explicit emotion-control input.

## Samples

Generated with the settings below (`cfg_scale=4.5`, `temperature=0.4`, `repetition_penalty=1.45`).

| Style | Prompt | Audio |
|---|---|---|
| Pronunciation | *Slovenščina je lep jezik s šumniki č, š in ž.* | <audio controls src="https://huggingface.co/texdata/Gepard-Slovenian-TTS/resolve/main/samples/01_sumniki.wav"></audio> |
| Neutral | *Danes je lep sončen dan in ptice pojejo na drevesih.* | <audio controls src="https://huggingface.co/texdata/Gepard-Slovenian-TTS/resolve/main/samples/02_nevtralno.wav"></audio> |
| Happy | *Kako čudovit dan, res sem srečen!* | <audio controls src="https://huggingface.co/texdata/Gepard-Slovenian-TTS/resolve/main/samples/03_veselo.wav"></audio> |
| Sad | *Žal mi je, vse je izgubljeno.* | <audio controls src="https://huggingface.co/texdata/Gepard-Slovenian-TTS/resolve/main/samples/04_zalostno.wav"></audio> |
| Angry | *Dovolj mi je tega, takoj nehaj!* | <audio controls src="https://huggingface.co/texdata/Gepard-Slovenian-TTS/resolve/main/samples/05_jezno.wav"></audio> |
| Question | *Kaj se je zgodilo, kje si bil ves ta čas?* | <audio controls src="https://huggingface.co/texdata/Gepard-Slovenian-TTS/resolve/main/samples/06_vprasanje.wav"></audio> |
| Surprise | *Neverjetno, kako je to sploh mogoče?* | <audio controls src="https://huggingface.co/texdata/Gepard-Slovenian-TTS/resolve/main/samples/07_presenecenje.wav"></audio> |
| Whisper | *Tiho, nekdo prihaja, ne povej nikomur.* | <audio controls src="https://huggingface.co/texdata/Gepard-Slovenian-TTS/resolve/main/samples/08_sepet.wav"></audio> |

## Generation settings (important)

Use these — they are the tuned sweet spot. **Prefix every input with `sl: `** (the language tag is
caller-owned; the model was trained on it).

```
temperature = 0.4        # clear articulation
cfg_scale   = 4.5        # strong text guidance → audible č/š/ž (2.5–4.5 usable; too high = harsh)
cfg_frames  = 25         # onset-only guidance
repetition_penalty = 1.45 # REQUIRED — without it, cfg 4.0 occasionally fails to terminate (runaway)
repetition_window  = 32
```

## Architecture / codec

- Backbone: `Qwen3_5ForCausalLM`, 14 layers, hidden 1024, vocab 248320, ~555M params, bf16.
- Codec: **NVIDIA NeMo NanoCodec** `nvidia/nemo-nano-codec-22khz-1.89kbps-21.5fps` — GroupFSQ, 8 layers,
  fsq_levels `[8, 7, 6, 6]`, 21.5 frames/s, 22050 Hz output.
- Language routing: transcripts prefixed `"{lang}: {text}"` before tokenization.

## How it was made

- **Data:** [ARTUR](http://hdl.handle.net/11356/1776) Slovenian read speech. Clean multi-speaker
  selection (speakers with ≥ 25 utterances; drops single-clip pseudo-speakers), **80000 utterances
  across ~1348 speakers**, 0.8–14 s, resampled to 22050 Hz.
- **Audio → tokens:** fused NanoCodec encode (slice/resample/encode in one GPU pass).
- **Training:** LoRA on the Qwen3.5 backbone (base frozen), **5 epochs**, single GPU, via
  [gepard-train](https://github.com/nineninesix-ai/gepard-train) (`text_layout=repeat_ft`, the
  short-register fix). Language tag `sl`, grapheme text (no G2P).

## Usage

Run with the [gepard-inference](https://github.com/nineninesix-ai/gepard-inference) runner + the NeMo
codec (needs `nemo_toolkit[tts]` for decoding):

```python
import torch, soundfile as sf
from gepard_inference import GepardRunner, UnfoldedCodecModel   # gepard-inference package

dev = "cuda"
runner = GepardRunner.from_checkpoint("texdata/Gepard-Slovenian-TTS", device=dev)
codec  = UnfoldedCodecModel.from_pretrained("nvidia/nemo-nano-codec-22khz-1.89kbps-21.5fps").eval().to(dev)

tokens = runner.generate(
    "sl: Danes je lep sončen dan in ptice pojejo.",   # MUST prefix "sl: "
    temperature=0.4, cfg_scale=4.5, cfg_frames=25,
    repetition_penalty=1.45, repetition_window=32, max_frames=800,
)
codes = tokens.unsqueeze(0).to(dev)
clen  = torch.tensor([codes.shape[-1]], device=dev)
audio, alen = codec.decode_from_codes(codes, clen)
sf.write("out.wav", audio[0, : int(alen[0])].cpu().numpy(), 22050)
```

## Limitations

- Overall quality is a first-pass; not production-grade. Best on short, declarative sentences.
- Grapheme-based: rare/foreign words and unusual stress may be mispronounced.
- Trained only on ARTUR read speech → neutral prosody, no emotion/expressive control tuned yet.
- Requires the generation settings above; low `cfg_scale` (< 1) is unstable, very high (≈ 9) gets harsh.

## References

- Base model: [nineninesix/gepard-1.0](https://huggingface.co/nineninesix/gepard-1.0) (Apache-2.0)
- Training: [gepard-train](https://github.com/nineninesix-ai/gepard-train) ·
  Inference: [gepard-inference](https://github.com/nineninesix-ai/gepard-inference)
- Codec: [nvidia/nemo-nano-codec-22khz-1.89kbps-21.5fps](https://huggingface.co/nvidia/nemo-nano-codec-22khz-1.89kbps-21.5fps) · [NVIDIA NeMo](https://github.com/NVIDIA/NeMo)
- Base pretraining data (gepard-1.0): LAION Emilia, EmoNet-Voice
- Fine-tune data: [ARTUR Slovenian speech corpus](http://hdl.handle.net/11356/1776) (CLARIN.SI)
- Tokenizer: `nineninesix/qwen3_5-full-attn-only-14`

## License

Apache-2.0 for the model weights (inherited from gepard-1.0). The NanoCodec used for decoding is under
NVIDIA's Open Model License.

## Citation

**Cite this work** (Tadej Fius, MediaAtlas Ltd):

```bibtex
@misc{fius2026gepard,
  title        = {Gepard Slovenian TTS},
  author       = {Fius, Tadej},
  year         = {2026},
  publisher    = {MediaAtlas Ltd},
  howpublished = {Hugging Face},
  url          = {https://huggingface.co/texdata/Gepard-Slovenian-TTS}
}
```

Upstream / source citations:

```bibtex
@misc{gepard, title={Gepard 1.0 (Qwen3.5-backbone TTS, NanoCodec)}, author={{nineninesix}}, url={https://huggingface.co/nineninesix/gepard-1.0}}
@misc{artur, title={ARTUR: Speech Database of Slovene}, author={{CLARIN.SI}}, url={http://hdl.handle.net/11356/1776}}
```
LoRA fine-tune of nineninesix/gepard-1.0 on the ARTUR read-speech corpus (CLARIN.SI).