Skip to content
Audio

OmniVoice-Africa

zero-shot TTS для 93 африканских языков и английского

1 min readupdated

OmniVoice-Africa — это TTS-модель от команды Ghana NLP / AfriSpeech, адаптация k2-fsa/OmniVoice. База — Qwen3-0.6B и 8-слойный дискретный кодек HiggsAudio; итоговый размер — около 0,81 млрд параметров. Модель синтезирует речь на 93 африканских языках и английском, включая Twi, Hausa, Yoruba, Swahili, Zulu, Igbo, Amharic и другие.

Ключевые особенности

  • 93 африканских языка и английский: от Twi и Hausa до Zulu, Igbo, Amharic и Shona.
  • Фонетическая обработка IPA через africa-g2p и espeak-ng снижает неоднозначность произношения.
  • Zero-shot клонирование голоса по референсному аудио 3–5 секунд.
  • Настроенный темп речи: speed=0.85, около 145 слов в минуту, для естественного звучания.
  • Обучение на 897 часах речи AfriSpeech (387 175 фраз).

Пример

import soundfile as sf
from omnivoice.models.omnivoice import OmniVoice
model = OmniVoice.from_pretrained('AfriSpeech/omnivoice-africa', device_map='cuda:0')
audios = model.generate(
    text='Akwaaba! Yɛfrɛ me Kofi, mepɛ sɛ mereka Twi kyerɛ wo nnɛ.',
    language='twi',
)
sf.write('twi_output.wav', audios[0], model.sampling_rate)
ПодробнееОбучение: 897 ч речи (387 175 фраз), LoRA RSLoRA r=32, α=64, merged в Qwen3-0.6B; лучший validation loss — 3.7184.
Железо: обучение на NVIDIA H200 (140 ГБ VRAM, bf16 + TF32); требования к VRAM для инференса в карточке не указаны.
Ограничения: для английского нужен espeak-ng; русский язык не поддерживается. Бенчмарков в карточке нет.
Discuss Telegram 0 98

Will it run?

Runs in F32

Fits without quantisation.

  • F32weights 1.7 GB · ready files2.4 GB
  • BF16weights 0.9 GB1.4 GB
  • FP8 / INT8recommendedweights 0.4 GB1.0 GB
  • 4-bitweights 0.2 GB0.8 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · F32 — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # AfriSpeech/omnivoice-africa
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download AfriSpeech/omnivoice-africa --local-dir ~/models/omnivoice-africa
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to omnivoice-africa
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

  1. Qwen3-0.6B-Basefamily root
  2. Qwen3-0.6B
  3. OmniVoice
  4. omnivoice-africa

New posts about versions of Qwen3-0.6B-Base (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 3All on Hugging Face

About the model

AfriSpeech/omnivoice-africa
Author
AfriSpeech
Category
Audio
Released
Posted

On NeyroLearn

Hugging Face

980

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments

Audio

OmniVoice

OmniVoice