Small files
Files up to 5 MB are passed through as provided, subject to the 50 MB upload limit.
Free Audio Deepfake Detector
AI voice detector for cloned-voice and synthetic-speech signals in an audio file.
Useful for voice-scam triage. Free, no signup, with clear limits.
Audio · Drop files or select multiple to begin bulk analysis
Audio
File size limit: 50MB (MP3, WAV, M4A)
Audio · Drop files or select multiple to begin bulk analysis
Audio
File size limit: 50MB (MP3, WAV, M4A)
Uploads Not Retained Current request only
Up to 50MB MP3, WAV, M4A
Clear Limits Not identity proof
Scope and limits
What happens to an audio upload
Larger files
The browser converts one representative segment of up to 30 seconds to 16 kHz mono before analysis.
Result meaning
The model reviews audible patterns such as breathing, prosody and noise continuity. It is not a laboratory-grade speaker-authentication test.
Phone codecs, aggressive noise removal and music can erase or imitate useful signals. For fraud, legal or identity decisions, verify the caller through a second channel and seek specialist review.
AI Voice Detector — Is This Audio Real or Cloned?
DeepFakeCheck reviews audible patterns that may be associated with cloned or synthetic speech. Upload an MP3, WAV or M4A file up to 50 MB. The result is a screening signal, not speaker-identity proof, and the upload is not retained after the current request.
How AI voice detection works
The current service sends audio to a multimodal analysis model with explicit forensic guidance. It does not claim a separate laboratory-grade spectrogram or speaker-authentication model. The review focuses on audible evidence such as:
- Breathing and mouth sounds: missing or repeated pauses can be a clue, but noise removal can create the same effect.
- Prosody and emotion: unusually flat or uniform delivery may be relevant only when supported by other evidence.
- Noise continuity: sudden digital silence or metallic sibilants may warrant review, while phone and platform codecs can produce false positives.
Which part of the file is analyzed
Files up to 5 MB are read as provided. For larger files, the browser converts one representative segment of up to 30 seconds to 16 kHz mono before analysis. If the suspicious phrase is outside that segment, trim the file around the relevant moment and analyze it again.
Common use cases for voice detection
- Voice scam protection. Voice-clone scams targeting families ("Dad, I'm in trouble, send money") are now widespread. Use the detector when you get a suspicious voice message or recorded call.
- Newsrooms and podcasters verifying audio clips of public figures before quoting them.
- Call centers and KYC teams screening voice authentication attempts.
- Voice actors and creators checking whether their voice has been cloned without permission.
- Court and insurance investigators triaging audio evidence before forensic review.
Voice detection FAQ
How long should the sample be? Use a segment long enough to include several complete phrases. Very short clips contain less evidence and should be treated with lower confidence.
Will it work on phone-quality audio? It can review the clip, but telephone codecs, compression and denoising remove useful detail and can imitate synthetic artifacts.
What if only one phrase is AI-generated? Whole-file analysis can miss a short inserted phrase. Trim around the suspected section and compare it with nearby speech when possible.
Can it detect AI music? The current guidance is designed for speech. Do not rely on it for singing, instrumental music or music-generation attribution.
Is my audio stored? No. The file is used only for the current analysis request and is not retained by DeepFakeCheck. We never train models on user submissions.