Skip to content
Audio

ishiki-deepfake-0.5

a compact audio deepfake detector

1 min readupdated

The ishiki-labs/ishiki-deepfake-0.5 model detects spoofed and synthetic voices (audio deepfakes) in audio streams. Built on the facebook/wav2vec2-xls-r-300m architecture, it has been pruned down to 86.75M parameters. The tool analyzes 16kHz audio recordings and outputs a numerical spoofing probability score. It helps developers and security researchers filter voice messages and protect systems against spoofing.

Key features

  • Based on the wav2vec2 backbone with a 72.5% reduction in parameter count
  • Runs in a single pass (One-Forward-Pass Optimization) without backpropagation
  • Benchmarked across multiple datasets, including ASVspoof 2019 and 2021
  • Weighs just 353.6 MB in .safetensors format

What you can do

  • Integrate it into call verification pipelines to defend against voice fraud
  • Screen audio files in bots or web services for synthetic voice generation
  • Test speech recognition systems for resilience against spoofing attacks

Example

import numpy as np
from huggingface_hub import snapshot_download
import sys

repo_dir = snapshot_download("ishiki-labs/ishiki-deepfake-0.5")
sys.path.insert(0, repo_dir)
from inference import create_predictor

predict = create_predictor(repo_dir, device="cuda:0", batch_size=32)
wavs_16khz = np.random.randn(4, 64600).astype(np.float32) * 0.2
scores = predict(wavs_16khz)
print("Spoof scores:", scores)
Read moreDetails
Parameters: 86.75M
ASVspoof 2019 LA EER: 0.1630%
ASVspoof 2021 LA EER: 1.5133%
Requirements: GPU (cuda:0) for fast batch processing
Discuss Telegram 1 17

Will it run?

Runs in F32

Fits without quantisation.

  • F32weights 0.3 GB · ready files0.9 GB
  • BF16weights 0.2 GB0.7 GB
  • FP8 / INT8recommendedweights 0.1 GB0.6 GB
  • 4-bitweights 0.0 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 · F32 — 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.

    # ishiki-labs/ishiki-deepfake-0.5
    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 ishiki-labs/ishiki-deepfake-0.5 --local-dir ~/models/ishiki-deepfake-0.5
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to ishiki-deepfake-0.5
  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 wav2vec2-xls-r-300m (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 9All on Hugging Face
Fine-tunes 12All on Hugging Face

About the model

ishiki-labs/ishiki-deepfake-0.5
Author
ishiki-labs
Category
Audio
Released
Posted

On NeyroLearn

Hugging Face

171

Ask AI about this model

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

Comments