Skip to content
Audio

VibeVoice-CSK1

ASR для чешского и словацкого с диаризацией и таймкодами

1 min read

Это полный BF16-checkpoint на базе microsoft/VibeVoice-ASR: в него влит QLoRA-адаптер, обученный на 450 часах чешской и словацкой речи. Модель на 8,67 млрд параметров распознаёт речь и выдаёт текст, метки спикеров и таймкоды. В комплекте оба speech encoder, tokenizer и processor.

Ключевые особенности

  • Специализация на чешском и словацком; задача — automatic-speech-recognition.
  • QLoRA-адаптер уже merged в VibeVoice-ASR, веса не квантованы — BF16.
  • На выходе не только текст, но и speaker labels с временными метками.
  • Готовый inference.py: можно передать локальную папку или repo ID, есть офлайн-режим через --local-files-only.
  • Greedy decoding; --max-new-tokens по умолчанию 4096, при неполной генерации сохраняется частичный вывод.

Пример

pip install huggingface-hub==0.36.2
hf download lofcz/VibeVoice-CSK1 --local-dir VibeVoice-CSK1
cd VibeVoice-CSK1
pip install -r requirements.txt
python inference.py recording.wav --device cuda:0 --output transcription.json
ПодробнееАдаптер: rank 64, alpha 32, checkpoint step 2564; экспорт — unquantized BF16.
Тестировали на 24 GB RTX A5000; нужны Python 3.11+, CUDA GPU и больше памяти для длинных записей.
Для форматов вне SoundFile нужен FFmpeg; длинные аудио лучше резать на клипы.
В merge_manifest.json записаны base revision, хэши адаптера, версии зависимостей и проверки тензоров.
Discuss Telegram 0 20

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 16 GB · ready files18 GB
  • FP8 / INT8recommendedweights 8.1 GB9.4 GB
  • 4-bitweights 4.5 GB5.5 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 3080 · BF16 — from RTX 4090

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.

    # lofcz/VibeVoice-CSK1
    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 lofcz/VibeVoice-CSK1 --local-dir ~/models/VibeVoice-CSK1
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to VibeVoice-CSK1
  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 VibeVoice-ASR (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 8All on Hugging Face
Merges 1All on Hugging Face

About the model

lofcz/VibeVoice-CSK1
Author
lofcz
Category
Audio
Released
Posted

On NeyroLearn

Hugging Face

200

Ask AI about this model

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

Comments

Audio

Swanada

Swanada