dots.tts-mf-1step

This repository provides a standalone dots.tts artifact for fixed one-step inference. The sampling contract is stored in config.json and selected automatically. Sampling options should be omitted.

Quick start

Install dots.tts from its official repository, then load this artifact exactly like a standard dots.tts model:

from dots_tts.runtime import DotsTtsRuntime
import soundfile as sf

runtime = DotsTtsRuntime.from_pretrained(
    "dots-studio/dots.tts-mf-1step",
    precision="bfloat16",
)

result = runtime.generate(
    text="Hello, this is a one-step synthesis test.",
    prompt_audio_path="reference.wav",
    prompt_text="The exact transcript of the reference audio.",
)
sf.write("output.wav", result["audio"].float().cpu().squeeze().numpy(), result["sample_rate"])

Scope and limitations

This model was optimized for the fixed one-step path. Multi-step quality is not claimed. A fresh public benchmark evaluation is not included in this artifact. High-fidelity voice cloning must be used only with authorization and consent; do not use it for impersonation, fraud, or disinformation.

Citation

@article{dotstts2026,
  title         = {dots.tts Technical Report},
  author        = {dots.tts Team},
  year          = {2026},
  eprint        = {2606.07080},
  archivePrefix = {arXiv},
  primaryClass  = {cs.SD},
}

License

Released under Apache-2.0.