Skip to content
Audio

ribt

голос для роботов с задержкой 17,6 мс

2 min read

ribt — английская TTS-модель на 54M параметров от Municipal Robotics. Она синтезирует речь потоково, чтобы робот успевал ответить, пока человек ещё рядом. В основе — цепочка lilt → acoustic (18,65M) → flow (21,88M) → Vocos (13,5M); все сети описаны в tensor IR tang и обучены без PyTorch.

Ключевые особенности

  • Только английский (US) и один голос; выход 24 kHz.
  • Первое аудио через 17,6 мс, RTF 0,027 на M4 Max (batch 1, warm, seed 0).
  • Один шаг flow matching вместо 16-шагового teacher — добавляет детали, которые дал бы teacher.
  • Одни и те же скомпилированные программы работают на Metal, CUDA, OpenCL, C и WebGPU.
  • Голос дистиллирован из Kokoro-82M af_heart: 135 ч рендера, весов Kokoro в модели нет.

Пример

W=ribt-weights
target/release/ribt-tts synth --run $W \
  --ckpt acoustic.safetensors --flow $W/flow.safetensors \
  --nfe 1 --shortcut --temp 0.3 \
  --voc $W/vocoder.safetensors --ref $W/ref.wav \
  --seed 0 --pitch 1 --accent minnesota \
  --text "Hi, I'm right here. What do you need?" --out-dir out/
ПодробнееБенчмарки на M4 Max: первое аудио 17,6 мс, RTF 0,027; UTMOS 4,39 против 4,52 у Kokoro af_heart; WER на held-out 50 — 1,73% у обеих.
Ограничения: только английский (US) и один голос; естественность немного уступает teacher.
Браузерной сборке нужен WebGPU и около 140 МБ загрузки при первом запуске.
Лицензия MIT; веса Vocos — MIT.
Discuss Telegram 0

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.1 GB · ready files0.6 GB
  • FP8 / INT8recommendedweights 0.0 GB0.5 GB
  • 4-bitweights 0.0 GB0.5 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 · BF16 — 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.

    # municipal-robotics/ribt
    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 municipal-robotics/ribt --local-dir ~/models/ribt
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to ribt
  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

New posts about versions of ribt (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

municipal-robotics/ribt
Author
municipal-robotics
Category
Audio
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

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

Comments