Skip to content
Audio

AADD v3

находит деградации речи каждые 20 мс

1 min readupdated

AADD v3 — модель автоматического обнаружения деградаций аудио (Automatic Audio Degradation Detection) от agertel. Это файнтюн Whisper-small encoder на 88 млн параметров: на 8-секундных окнах она выдаёт вероятности 12 типов деградаций для каждого 20-мс фрейма. Предназначена для чистки ASR/TTS-датасетов и аудита записей.

Ключевые особенности

  • 12 классов: шум, музыка, реверберация, клиппинг, low-pass / narrow band, high-pass, кодеки, packet loss, mains hum, клики, pitch shift и time stretch.
  • Несколько деградаций могут присутствовать одновременно; выход — sigmoid на каждый класс на 20-мс фрейм.
  • Основа — Whisper-small encoder; чекпоинт весит 177 МБ и загружается через Detector("v3").
  • Обучена на английском и русском; есть CLI для сканирования датасета и HTML-отчёта.
  • Для реальных записей рекомендуется v3; v1 и v2 — лёгкие CPU-модели для 7 синтетических классов.

Пример

from aadd import Detector

det = Detector("v3", resolution=1.0)
res = det.predict_file("call.wav")
res.detected
res.frame_probs.shape
ПодробнееROC-AUC на VoiceBank-DEMAND для шума — 0.985; frame macro-AP на held-out синтетике — 0.955 (EN) и 0.959 (RU).
Чекпоинт 177 МБ, backbone 88M параметров; обучение заняло 2.4 ч на одной H100.
Ограничения: деградации синтетические, реальные тесты покрывают шум и реверберацию; модель рассчитана на читаемую речь и вход 16 кГц.
Видео-пример
Discuss Telegram 1

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.

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

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

About the model

agertel/aadd
Author
agertel
Category
Audio
Released
Posted

On NeyroLearn

Hugging Face

01

Ask AI about this model

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

Comments