40-person blind test: people spot AI video at 52%, barely better than a coin flip
Resident-Anxiety6427 · reddit · 2026-10-27
A Redditor ran an informal blind test with 40 friends, coworkers, and his mom's book club: 10 clips (5 real iPhone shots, 5 Dreamina-generated versions of the same scenes, prompted via ChatGPT), same color grade, no sound.
- Average accuracy: 52% — a coin flip.
- The kneading-dough clip was easiest to catch (86% spotted it; hands gave it away).
- The night street scene fooled almost everyone: 74% judged the AI version real.
- Under-30s did slightly worse than over-50s.
The most striking finding was overconfidence: many participants said "obviously that one" and were wrong. The author notes it's not scientific — just a dinner argument gone out of hand — and asks if people want a bigger version.
More from Multimodal
- Uncensored Wan 1.3B fine-tune hits Hugging Face, runs video gen on 8GB VRAM — Promptmethus · 2026-10-27
- Illustration-to-real tiger headdress recreation wows with near-perfect detail match — gizakdag · 2026-10-27
- Gradium's open-weight TTS runs offline on iOS/macOS, benchmarked vs Kokoro and PocketTTS — amos_gyamfi · 2026-10-27
- SATE-Bench: 316 scenes show MLLMs stop exploring too early with high confidence — Tianchen Guo · 2026-10-27
- Two-year-old patch still fixes ComfyUI's slow checkpoint loading: 225MB/s to 2.5GB/s — Solongtomegrandma · 2026-10-27
- A lava-slide first-person video made with Dreamina for just 29 cents — Certain_Menu_8412 · 2026-10-27