LocalClaw/TTS/VeriSpoof-Base
Local audio-language model

VeriSpoof-Base

Catalogue summary: Local audio-language model for speech deepfake detection, source comparison and explanations of recording conditions. Combines modified MiniCPM5-2B and Qwen3-ASR audio-encoder weights with an audio tokenizer and localization head. Returns text and scores for research use.

Repository editorial metadata; verify comparative claims in the linked upstream material.

GPU recommendedspeech deepfake detection and source analysis1 languagesCC-BY-NC-4.0; upstream MiniCPM5 and Qwen3-ASR components Apache 2.0
Choose an app

Start with the recommended app, or choose another supported option.

Compare speech models
Open model filesModel weights and official local setupOpen in UnslothLaunches Unsloth Desktop on this modelOpen on Hugging FaceFiles, licence and model cardCheck my machineOptional: verify the fit with your saved hardware

Desktop app links require the app to be installed. If nothing opens, LocalClaw will show app-download and model-file fallbacks.

Catalogue quality
Not rated
Catalogue speed
Not rated
Model size
2.9B parameters (author); 6.362 GB safetensors
Voices
N/A: text and detection scores; no speech synthesis, cloning or streaming

Can VeriSpoof-Base run locally?

VeriSpoof-Base can run locally for speech deepfake detection and source analysis over the first 8 seconds of a recording. Use the verified setup options on this page; no terminal command is required to choose the right path.

CC-BY-NC-4.0; upstream MiniCPM5 and Qwen3-ASR components Apache 2.0 license. Review upstream restrictions before commercial use.

Upstream source
audio-languageaudio-forensics

Model details

Type
Local audio-language model
Family
verispoof
Latency
not rated
Formats
pytorchtransformerssafetensors
Languages
en
Context
Official model card and bundled code/README.md specify about 12 GB CUDA VRAM, not a measured universal minimum or system-RAM floor. The default PyTorch API loads the released checkpoint locally. Input is resampled to 16 kHz; only the first 8 seconds are analyzed, and shorter clips are repeated to fill the window. English prompts and answers are documented; the multilingual tag does not establish a supported speech-language list. Optional vLLM accelerates reasoning only; detection and localization use PyTorch. Non-commercial research release; do not treat scores as definitive authenticity proof. No local inference test performed. Source: huggingface.co/AustinXiao/VeriSpoof-Base, including NOTICE and code/verispoof/inference.py.

Install locally

01
Check runtimeConfirm the backend supports pytorch, transformers, safetensors on your machine.
02
Open recommended setupUse the app and model links above. LocalClaw does not expose a terminal command.
03
Test locallyRun a short private audio prompt before moving into production workflows.

Good for

  • speech deepfake detection and source analysis
  • GPU recommended local workflows
  • audio-language, audio-forensics

Watch before shipping

  • Validate pronunciation, latency and artifacts with your own voice samples.
  • Review the upstream license and acceptable-use notes.
  • Benchmark on your target CPU, Apple Silicon or GPU setup.

Related TTS and speech models

Yang Xiao VeriSpoof-Plus Local audio-language model · Not rated Kaiyang Li Mizar-159M Local audio-language model · Q 8.7 · Speed 8.8 OpenLLM France / LINAGORA Luciole 8B Audio 1.0 Local audio-language model · Q 9.1 · Speed 6.7 jamiepine / Community Voicebox Local speech app · Q 9 · Speed 9.5 IndexTeam / Bilibili Index-Echo S2ST 2B Local speech translation model · Q 9 · Speed 7.8 SILVERTHRONE Ossicle-4B Local audio-language model · Not rated hexgrad Kokoro TTS Local TTS model · Q 9.2 · Speed 9.8 Speech Research (SWivid) F5-TTS v1 Base Local TTS model · Q 9.5 · Speed 9.2
CompareBrowse all TTS models Local AIBrowse LLM models macOS appGet LocalClaw