Skip to content
Other

HTDemucs Engine

4 стема синтетических звуков двигателя

1 min readupdated

Кастомная HTDemucs от Super3dMax, обученная с нуля разделять синтетические звуки двигателя и машин на четыре стема: whine, hiss, click и bump. Архитектура — гибрид спектрограммы и waveform-трансформера Demucs, 42 млн параметров.

Ключевые особенности

  • Разделяет смесь на тональный вой, широкополосный шип, короткие клики и низкочастотные удары.
  • HTDemucs: channels=48, bottom_channels=512, t_layers=5, nfft=4096, CaC, без Wiener-итераций.
  • Обучение: 1 024 синтетических трека + 128 валидационных, 60 000 шагов, EMA-веса.
  • Два самодостаточных GGUF для audio.cpp: f16 (~84 МБ, рекомендуется) и q8_0 (~60 МБ).
  • Также запускается через PyTorch/mss-training; GGUF совместим с узлом HTDemucs в SpectroCraft DAW.

Пример

audiocpp_cli --task sep --family htdemucs --model htdemucs-whine-hiss-click-bump_v1_f16.gguf --backend cpu --audio input.wav --out-dir stems_out

На вход нужен stereo 44.1 kHz WAV; на выходе — четыре WAV той же длины.

ПодробнееSDR на валидации: ALL +15.3 dB (SI-SDR +14.8), whine +22.4, click +5.1.
Обучение заняло ~15 ч на одной RTX 5070 Ti (16 ГБ).
Ограничения: только синтетические данные; клики короче STFT-фрейма и восстанавливаются частично; стерео L=R, частота 44.1 кГц.
Лицензия MIT.
Discuss Telegram 0 25

Will it run?

Runs in GGUF F16

The most precise of the ready quantised files fits.

  • GGUF F16weights 0.1 GB · ready files0.6 GB
  • GGUF Q8_0recommendedweights 0.1 GB · ready files0.6 GB

Tip: go 8-bit.GGUF Q8_0 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: GGUF Q8_0 — from RTX 2060 · GGUF F16 — 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.

    # Super3dMax/htdemucs-whine-hiss-click-bump_v1
    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 Super3dMax/htdemucs-whine-hiss-click-bump_v1 --local-dir ~/models/htdemucs-whine-hiss-click-bump_v1
    # The model is in ~/models. How to run it is in the model card (link above).
  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 htdemucs-whine-hiss-click-bump_v1 (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

Super3dMax/htdemucs-whine-hiss-click-bump_v1
Author
Super3dMax
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

250

Ask AI about this model

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

Comments