Skip to content
Other

Pronunciada

on-device pronunciation scoring with cloud-level accuracy

2 min readupdated

The Pronunciada library scores pronunciation from an audio recording and reference text, returning acceptance probabilities for phonemes, words, and sentences. Built on the ZIPA phoneme recognizer model, it adds lightweight adapters, calibrators, and prosody models across 10 languages, including Russian. The tool runs locally on-device, processing a short clip in tens of milliseconds, making it ideal for developers building educational platforms and language-learning apps.

Key features

  • Supports 10 languages, including English, Russian, Chinese, and Japanese
  • Runs entirely on-device with no cloud API calls required
  • Fast inference at around 50 milliseconds per clip on an Apple M3 Pro
  • Low memory footprint of roughly 0.3 GB using the quantized base model

What you can do

  • Integrate pronunciation scoring into mobile or desktop language-learning apps
  • Check pronunciation accuracy across individual phonemes, words, and full sentences
  • Use the CLI tool for quick batch checks on audio files

Example

from pronunciada import Scorer

scorer = Scorer("en")
result = scorer.score("clip.wav", "It was a beautiful day for a float")
for word in result.words:
    print(word.text, round(word.prob, 2))
Read moreDetails
Benchmarks: On the speechocean762 dataset, correlation with human raters hits 0.685 for phonemes, 0.609 for words, and 0.748 for sentence-level accuracy
Hardware requirements: Around 0.3 GB of VRAM/RAM with the quantized ZIPA model (67 MB)
Limitations: Accuracy on non-native learner speech has only been evaluated for English and Chinese; requires reference text; no iOS, Android, or WebAssembly builds available yet
Discuss Telegram 0

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.0 GB · ready files0.5 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.

    # kimitonatsu/pronunciada
    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 kimitonatsu/pronunciada --local-dir ~/models/pronunciada
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to pronunciada
  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 zipa-small-crctc-500k (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 2All on Hugging Face
LoRAs and adapters 1All on Hugging Face

About the model

kimitonatsu/pronunciada
Author
kimitonatsu
Category
Other
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