Deepfake Word Detection by Next-token Prediction using Fine-tuned Whisper

2026-02-26

AI Summary

This study addresses the detection of synthetic replacement words (deepfake words) in speech by proposing a novel approach based on fine-tuning the Whisper model. Leveraging Whisper’s next-token prediction mechanism during transcription, the method identifies anomalous synthetic tokens and incorporates partially vocoder-processed speech data for augmentation, substantially reducing annotation costs. This work is the first to apply Whisper’s next-token prediction capability to deepfake word detection, achieving simultaneously low transcription and detection error rates on in-domain data. On out-of-domain data, the approach matches the performance of specialized ResNet-based detection models, demonstrating strong practical potential, although room remains for improving generalization.

Technology Category

• Natural Language Processing: Speech
• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Diffusion Models for Vision

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Deepfake speech utterances can be forged by replacing one or more words in a bona fide utterance with semantically different words synthesized by speech generative models. While a dedicated synthetic word detector could be developed, we investigate a cost-effective method that fine-tunes a pre-trained Whisper model to detect synthetic words while transcribing the input utterance via next-token prediction. We further investigate using partially vocoded utterances as the fine-tuning data, thereby reducing the cost of data collection. Our experiments demonstrate that, on in-domain test data, the fine-tuned Whisper yields low synthetic-word detection error rates and transcription error rates. On out-of-domain test data with synthetic words produced by unseen speech generative models, the fine-tuned Whisper remains on par with a dedicated ResNet-based detection model; however, the overall performance degradation calls for strategies to improve its generalization capability.

Problem

• Deepfake
• Synthetic word detection
• Speech forgery
• Next-token prediction
• Generalization

Innovation

• next-token prediction
• fine-tuned Whisper
• deepfake word detection
• partially vocoded utterances
• speech generative models