Volver
Gemini Omni 1.2Gemini Omni 1.1Video generationMultilingualConsistency

I Tested Gemini Omni 1.2 vs 1.1 Across Three Indian Regional Languages

Harsh Sinha⭐⭐⭐Sitio del autorNov 21, 2026Visitar Gemini
100% Útil0% No útil
0 Comentarios

Model

Gemini Omni 1.2 vs 1.1

What I did

I compared six 360p AI-generated video ads across Kannada, Gujarati, and Marathi, testing one pair per language. I evaluated visuals, audio, and transcription against 12 criteria, used blind A/B judgments in both orders, and checked evaluation reliability with blur, freeze, and repeated-audio controls.

What happened

1.2 beat 1.1 in Kannada (88.2 vs 83.9) and Marathi (100 vs 90.7), but lost in Gujarati (95.0 vs 95.7). Mean gain: 4.3 points across the three pairs.

Worth it?

Yes, cautiously. Promising for multilingual ads, but one pair per language cannot prove a consistent lead.

Screenshot 2026-11-21 at 6.30.21 PM.png

Comentarios0

Inicia sesión para comentar
  • Aún no hay comentarios.

Artículos relacionados