---
title: OmniVoice Word-Control
emoji: 🎛️
colorFrom: indigo
colorTo: pink
sdk: gradio
sdk_version: 6.10.0
app_file: app.py
pinned: false
license: cc-by-nc-4.0
short_description: Voice cloning with word-level prosody control (WordVoice-5A)
models:
  - multimodalart/omnivoice-word-control
  - k2-fsa/OmniVoice
---

# OmniVoice Word-Control

Zero-shot voice cloning with explicit **word-level control** over duration, boundary/pauses,
energy, pitch, and tone contour — the [WordVoice](https://huggingface.co/papers/2607.06461)
task realized on [OmniVoice](https://huggingface.co/k2-fsa/OmniVoice)'s masked-diffusion LM
via inline control tokens, fine-tuned on
[WordVoice-5A](https://huggingface.co/datasets/XXH333/WordVoice-5A) (English, ~2,138h).

UI and inline-tag syntax adapted from
[hugging-apps/wordvoice-tts](https://huggingface.co/spaces/hugging-apps/wordvoice-tts);
generation flow adapted from the official
[k2-fsa/OmniVoice](https://huggingface.co/spaces/k2-fsa/OmniVoice) Space.
Example reference clip from the WordVoice-5A test split (CC-BY-4.0).