Audio, Video and Image Deepfake Detection Benchmarks

Model

DETECT-World ranked #1 across audio, image, and video detection

DETECT-World is the first deepfake detection model built on a world model architecture with the highest accuracy on third-party benchmarks.

99.5%

Pooled detection accuracy, Podonos Audio DFD Bench

Last updated August 2026

IMAGE

95.8%

Average accuracy on images

False negative rate of 4.3% and false positive rate of 4% when benchmarked against 1,000 image files.

VIDEO

98.2%

Average accuracy on video

False negative rate of 2% and false positive rate of 0.95% when benchmarked against 1,000 video files.

RTF

0.12

Real-time factor

shows ultralow latency that provides results almost instantly.

Deepfake detection benchmarks

RESEMBLE AI 

COMPETITOR

Audio deepfake detection — Accuracy

Pooled accuracy across all audio files. Higher is better.

Resemble DETECT-World

99.5%

Whispeak

97.7%

Aurigin AI

96.8%

Pella Research

95.8%

Pindrop

95%

Accuracy % (higher = better)

100%

Audio DFD Benchmark · Podonos · August 2026 · Source

Audio deepfake detection — False Negative Rate (FNR)

Misidentification of AI files as authentic. Lower is better.

Resemble DETECT-World

0.4%

Whispeak

1.7%

Pella Research

2.8%

Reality Defender

3.6%

Pindrop

3.7%

False negative rate (lower = better)

Test sets include mp3, wav, flac, ogg, m4a, webm across 4,524 files. Source

LANGUAGE COVERAGE

Over 50 languages detected

ALSO SUPPORTED BY CHATTERBOX TTS

Arabic

Chinese

Danish

Dutch

English

Finnish

French

German

Greek

Hebrew

Hindi

Italian

Japanese

Korean

Malay

Norwegian

Polish

Portuguese

Russian

Spanish

Swahili

Swedish

Turkish

Ukranian

Vietnamese

Thai

Indonesian

Romanian

Czech

Slovak

+ over 20 additional languages

Validated against MLAADv10. Detection relies on generation artifact patterns, not language-specific features — enabling generalization to languages not seen during training. EER and accuracy figures from the Hugging Face Speech DF Arena and DFBench Speech/Image leaderboards, March 2026. Resemble AI does not control test set composition. Image figures from DFBench Image 2025. Video EER from internal evaluations on held-out test sets.

VERIFY

PerTh watermarker: survives compression, re-encoding, and attack

Detection accuracy across 18 real-world attack conditions. PerTh V2 ships with improved robustness across reverb, pitch shift, and spectral manipulation.

~100%

Detection on clean and compressed audio

PerTh V2 · No-attack + standard codecs

STRONG (>90%)

MODERATE (70-90%)

WEAK (<70%)

PerTh — Open Source, Audio Only

wav_dither_attack

100%

random_wav_wavelet_noise

90%

random_wav_reverb_attack

95%

random_wav_resample_attack

100%

random_wav_precision_attack

100%

random_wav_pitch_shift_attack

10%

random_wav_mulaw_attack

100%

random_wav_high_pass_attack

100%

random_wav_gaussian_noise_clipped

45%

random_spec_time_mask

100%

random_spec_stretch

100%

random_spec_scale

100%

random_spec_lowclip

100%

random_spec_highclip

100%

random_spec_gaussian_noise_clipped

100%

random_spec_contiguous_band_mask

72%

no_watermark

100%

no_attack

100%

Accuracy

100%

PerTh Multimodal

wav_dither_attack

100%

random_wav_wavelet_noise

100%

random_wav_reverb_attack

88%

random_wav_resample_attack

100%

random_wav_precision_attack

100%

random_wav_pitch_shift_attack

94%

random_wav_mulaw_attack

100%

random_wav_high_pass_attack

100%

random_wav_gaussian_noise_clipped

100%

random_spec_time_mask

100%

random_spec_stretch

98%

random_spec_scale

100%

random_spec_lowclip

100%

random_spec_highclip

100%

random_spec_gaussian_noise_clipped

100%

random_spec_contiguous_band_mask

98%

no_watermark

100%

no_attack

100%

Accuracy

100%

Watermark robustness = detection accuracy after each attack transform (1.0 = 100%). "no_attack" = clean audio. "no_watermark" = false positive rate on unwatermarked audio. PerTh V2 figures are pre-release internal evaluations.

DEPLOYMENT / INFRA

Deploy on the infrastructure that meets your needs.

Resemble AI models — DETECT-3B Omni and PerTh run on cloud API, on-prem, or air-gapped. Pick the deployment model that fits your security posture and latency requirements.

CLOUD API



Managed service

RESTful API, 99.9% uptime SLA, auto-scaling. Fastest time to value.

ON-PREMISES



Your infrastructure

Docker / Kubernetes. No data leaves your network. NVIDIA GPU optimized.

AIR-GAPPED



Zero internet dependency

Run Resemble AI in your own AWS, GCP, or Azure VPC. Isolated compute, private networking, and compliance with your cloud security policies.