Deepfake detectionVoice: speakers and sound

AASIST

The baseline open model against voice spoofing: it listens to the raw recording and tells a live person from synthesis or a replay. It errs in both directions - its output is a reason for a human to check, not proof.

The last open version came out in Oct 2021. The family has not been updated for a long time: the model still works, but do not expect fixes or new sizes.

Developer
NAVER Clova AI Research and EURECOM, South Korea
First release
Oct 2021
Latest release
Oct 2021
Sizes
weight files of 0.4 and 1.3 MB
License
Commercial use allowedMIT
Running
On your own serverAlso runs without a GPU
Industries
Security, Finance, Customer support, Public sector
Deploy on your server GitHub

What it does

  • Voice check during phone authentication
  • Filtering replays and synthesis in a voice menu
  • A baseline when comparing voice detectors
  • Fine-tuning for your own communication channel

Where it is used

Call centers and supportBanks and insuranceInformation securityPublic sector

Hardware requirements

LaptopLaptop or regular PC, up to 8 GB of VRAM — smaller versions
fits
1 GPUOne GPU with 16–80 GB — mid-size versions
no versions
ClusterServer with several GPUs — flagship versions
no versions

Versions

  1. AASIST-L (облегчённая)
  2. AASIST

How to run it

On your own serverWeights are downloaded from Hugging Face and served with vLLM (load and API) or llama.cpp (modest hardware). The model then runs in a closed network with no per-request fees.
How much hardware you needCalculate the VRAM for the model size, context length and number of concurrent requests.Open the hardware calculator

I can set this up end to end: pick the model size, deploy it on your server and connect it to your systems.

Frequently asked questions

Can AASIST be used in a commercial project?

Yes. License: MIT. It allows commercial use, but it is still worth having a lawyer review the license before launch.

What hardware does AASIST need?

At minimum: Laptop or regular PC, up to 8 GB of VRAM — smaller versions. Some versions also run on an ordinary CPU, without a GPU. You can calculate the exact VRAM for your model size and context in the hardware calculator.

Does AASIST support Russian?

Language does not matter for this model: it does not work with text.

Where can I download AASIST and what does it cost?

The AASIST weights are open and free to download. You only pay for the hardware it runs on and for the setup. Source links are at the bottom of this page.

How I deploy it for clients

  1. SelectionI pick the model size for your task and hardware and test it on your examples.
  2. DeploymentI deploy it on your server or in a closed network and provide an API.
  3. Fine-tuningI fine-tune it on your data (LoRA) or connect a knowledge base — whichever is cheaper for the task.
  4. IntegrationI connect it to your CRM, ERP, bot, website or team chat and set up monitoring.

Similar models

Deepfake detectionSSL Anti-spoofing (wav2vec 2.0 + AASIST)EURECOM · FranceCommercial use allowed

A step beyond AASIST: instead of raw audio it uses the wav2vec 2.0 speech encoder, which helps it hold up on unfamiliar synthesis methods. It errs in both directions - a human reviews the result.

DetailsDeepfake detectionAntiDeepfake (NII)National Institute of Informatics, Yamagishi Lab · JapanNon-commercial only

Seven speech encoders (wav2vec 2.0, XLS-R, MMS, HuBERT) post-trained to tell live speech from synthetic. The authors note themselves that quality depends heavily on the dataset; a human reviews the output.

DetailsVoice: speakers and soundWeSpeakerWeNet community · ChinaCommercial use allowed

A set of ready-made voiceprint models: checks whether the same person speaks in two recordings and helps split a recording by speaker. One of the models is built into pyannote 3.x.

Details

Source: github.com/clovaai/aasist. Data checked against the model card on 22 Sep 2026. Have a lawyer review the license before commercial launch.