General Knowledge
Spot the AI Voice Quiz
Deepfakes, Synthesia, ElevenLabs — can you tell real voices from AI?
Deepfakes, Synthesia, ElevenLabs — can you tell real voices from AI?
An AI can clone your voice from just 3 seconds of audio — and ElevenLabs' system generates speech with only 300 milliseconds of latency. This 50-question deep dive covers the history of text-to-speech from 1980s Kurzweil to WaveNet and Transformers, today's cloning tools including ElevenLabs, OpenAI Voice Engine, and Meta Voicebox, infamous scams like the $25M Hong Kong deepfake and the fake Biden robocall, detection techniques, watermark systems like SynthID, and the SAG-AFTRA and NO FAKES Act regulatory response.
Each round presents 10 randomized questions from a pool of 50, with four multiple-choice options and instant feedback after every answer. Your final score comes with a performance tier and shareable results.
You'll explore concatenative, parametric, and neural TTS, landmark systems like WaveNet and Tacotron, ElevenLabs' founding and funding, voice-cloning scams that cost victims $4 billion+ in 2023, detection tools like Hive and Resemble Detect, Google SynthID watermarking, the Scarlett Johansson vs OpenAI 'Sky' controversy, Tennessee's ELVIS Act, and the telltale signs that distinguish synthetic speech from a real human.
Modern tools like ElevenLabs can produce a recognizable voice clone from just 3 to 10 seconds of reference audio, and OpenAI's Voice Engine (announced March 2024) uses a 15-second reference sample.
A deepfake is AI-generated or AI-manipulated audio, image, or video that convincingly imitates a real person. Audio deepfakes use voice cloning to impersonate politicians, executives, or family members in scams.
Listen for unnaturally consistent tone, missing breath and swallow sounds, mispronounced proper nouns, mechanical pauses or 'um/uh,' and incorrect emotional inflection on complex sentences. Detection tools like Hive and Resemble Detect claim 90%+ accuracy.
Last updated: April 2026