Audio8-TTS

Supported Languages

Model Details

Audio8 TTS uses a DualAR architecture inspired by Fish Audio S2 Pro. The slow AR transformer predicts one semantic token for each audio frame. The fast AR transformer predicts the frame's codec codebooks, conditioned on the slow hidden state and preceding codebooks.

The bundled codec handles both reference-audio encoding and waveform decoding, so no additional codec checkpoint is required.

Installation

Python 3.10 or newer and a CUDA-capable GPU are recommended.

pip install "torch>=2.5.0" "torchaudio>=2.5.0" \
  "transformers>=4.57.0,=0.12" "safetensors>=0.4"

Usage

The model uses custom Transformers code. Review the files in this repository, then load it with trust_remote_code=True.

Zero-shot voice cloning

The reference transcript must match the spoken content in the reference audio.

import soundfile as sf
import torch
from transformers import AutoModel, AutoProcessor

model_id = "Audio8/Audio8-TTS-Preview-0.6b"
device = "cuda" if torch.cuda.is_available() else "cpu"
dtype = torch.bfloat16 if device == "cuda" else torch.float32

processor = AutoProcessor.from_pretrained(model_id, trust_remote_code=True)
model = AutoModel.from_pretrained(
    model_id,
    trust_remote_code=True,
    dtype=dtype,
).eval().to(device)

inputs = processor(
    text=["Welcome to Audio8 TTS."],
    reference_audio=["reference.wav"],
    reference_text=["The exact transcript of the reference recording."],
    return_tensors="pt",
)
inputs = {name: value.to(device) for name, value in inputs.items()}

with torch.inference_mode():
    output = model.generate(
        **inputs,
        max_new_tokens=1024,
        temperature=0.8,
        top_p=0.95,
        top_k=50,
        do_sample=True,
        return_dict_in_generate=True,
    )
    waveforms, waveform_lengths = model.decode_audio(output.codes)

audio = waveforms[0, : int(waveform_lengths[0])].float().cpu().numpy()
sf.write("output.wav", audio, model.config.codec_sample_rate)

Generation without a reference

Omit reference_audio and reference_text when a cloned voice is not needed:

inputs = processor(
    text=["This utterance does not use a reference voice."],
    return_tensors="pt",
)

For command-line inference, batching, and supervised fine-tuning, see the Audio8 TTS repository.

Deployment Options

CPU deployment: ONNX INT4

Audio8-TTS-Preview-0.6B-ONNX-INT4 packages Audio8 TTS for low-resource CPU inference with ONNX Runtime. Slow and Fast AR weights use weight-only INT4, while activations, KV caches, and the neural audio codec use FP16.

Normal synthesis loads only the Slow AR, Fast AR, and codec decoder sessions. Voice registration releases those sessions before loading the optional codec encoder, keeping peak memory controlled.

Get the ONNX INT4 model and follow the CPU ONNX Runtime guide.

High-throughput serving: SGLang Omni

Audio8 TTS now includes an SGLang Omni adapter for production-oriented GPU serving. It is installed as an independent model plugin and does not overwrite SGLang Omni core files.

The released adapter is validated against a pinned SGLang Omni revision and supports both generation without a reference and zero-shot voice cloning. See the SGLang Omni deployment guide for tested versions, installation, server configuration, and API examples.

Evaluation

Audio8 TTS Preview is the smallest model in this comparison at just 0.6B parameters. Despite using only a fraction of the parameters of the other systems, it delivers results in the first tier of industry-leading SOTA TTS models on the benchmarks below. In particular, it achieves th