Skip to content
Audio

CzechSpeech

чешская речь в текст на устройстве

2 min read

CzechSpeech — модель распознавания чешской речи от Moravian AI на базе Whisper large-v3-turbo. Это файнтюн на 0,81 B параметров, который транскрибирует парламентскую, читаную и зашумлённую речь точнее открытых моделей сопоставимого размера. В 4-битном кванте модель занимает 452 MiB и работает офлайн на ноутбуке или обычном телефоне.

Ключевые особенности

  • WER 1,8% на парламентском тесте, 7,3% на FLEURS и 6,5% на шумных сегментах.
  • Есть файлы для transformers (16-bit) и whisper.cpp: q5_0 — 547 MiB, q4_0 — 452 MiB.
  • На Pixel 9a q4_0 обрабатывает 6,9 с речи за 2,7 с, пик RAM — 0,83 ГБ.
  • На ноутбуке с RTX 4050 q5_0 занимает 0,89 ГБ GPU-памяти и работает примерно в 25 раз быстрее реального времени.
  • Обучена на 1 503 ч чешской речи, включая ParCzech4Speech; оценка без спикеров из теста и валидации.

Пример

import torch
from transformers import WhisperForConditionalGeneration, WhisperProcessor

processor = WhisperProcessor.from_pretrained("moravianai/CzechSpeech")
model = WhisperForConditionalGeneration.from_pretrained("moravianai/CzechSpeech", dtype=torch.float16).to("cuda")
features = processor(audio, sampling_rate=16000, return_tensors="pt").input_features.to("cuda", torch.float16)
ids = model.generate(features, language="cs", task="transcribe")
print(processor.batch_decode(ids, skip_special_tokens=True)[0])
ПодробнееWER: парламент — 1,8%, FLEURS — 7,3%, шум — 6,5%; у базовой Whisper large-v3-turbo — 9,5 / 11,4 / 29,5%.
4-битный q4_0 занимает 452 MiB; на Pixel 9a пик RAM — 0,83 ГБ, на RTX 4050 q5_0 — 0,89 ГБ GPU-памяти.
Ограничения: в основном формальная парламентская речь; нет таймстемпов, перевода и диаризации; сегменты длиннее 30 с нужно делить.
Лицензия двойная: бесплатно для некоммерческого использования по CC BY-NC 4.0, коммерческая — по запросу в Moravian AI.
Discuss Telegram 0 30

Will it run?

Working out the memory needs

The bot is still collecting this model's weight sizes; the estimate will appear here in a few minutes. Meanwhile you can rent a GPU with headroom.

Massed ComputeOpen Massed Compute

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.

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

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

About the model

moravianai/CzechSpeech
Author
moravianai
Category
Audio
Released
Posted

On NeyroLearn

Hugging Face

300

Ask AI about this model

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

Comments

Audio

AADD v3

AADD v3