Skip to content
Other

vyralsampler-models-nc

ONNX speech aligners for European languages

2 min read

A set of four ONNX-exported models based on wav2vec2 (from the VoxPopuli corpus via torchaudio), created by author JensLantz for the VyralSampler application. The models solve the task of forced alignment — determining the exact time boundaries of each spoken or sung word in audio and video. Unlike the original versions, these files work directly via ONNX Runtime without needing to install the heavy PyTorch framework. The tool will be useful for developers of audio and video editors, subtitle generators, and sample slicing systems.

Key features

  • Support for four languages: French (aligner_fr.onnx), German (aligner_de.onnx), Spanish (aligner_es.onnx), and Italian (aligner_it.onnx).
  • Running without PyTorch: the models are converted to ONNX (opset 17) and contain character labels directly inside the file metadata.
  • High conversion accuracy: the discrepancy of output values from the original on Torch does not exceed 5e-5, and the word boundary markup matches the WhisperX 3.8.6 algorithm.

What you can do

  • Integrate into an application or pipeline for automatic slicing of voice and vocal samples from videos.
  • Use in combination with speech recognition systems (similar to WhisperX) to create accurate word-by-word subtitles without heavy dependencies.

Example

In the VyralSampler application, the models are loaded into ONNX Runtime: the audio track from a video with speech or vocals in French, German, Spanish, or Italian is fed to the input, and the program receives the exact start and end timings of each word to slice the fragment.

Read moreDetails
Base models: VOXPOPULI_ASR_BASE_10K (FR, DE, ES, IT) from torchaudio
Export: legacy exporter from PyTorch 2.8 (opset 17), weights unchanged
License: CC BY-NC 4.0 (non-commercial use only)
Discuss Telegram 0

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.4 GB · ready files0.9 GB
  • FP8 / INT8recommendedweights 0.2 GB0.7 GB
  • 4-bitweights 0.1 GB0.6 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.

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

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

About the model

JensLantz/vyralsampler-models-nc
Author
JensLantz
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