Side-by-Side Clips: My Veo 3 vs 3.1 Test
I gave the same five prompts to both versions across two evenings, and my notebook ended up with a clear split. Talking scenes belong to version 3.1 in my runs: lip movement tracked my scripted lines more often, and the background audio stopped doing that odd underwater swirl right when a character spoke. The strange stuff is closer than I expected. When I asked for an origami crane folding itself on a desk, version 3 sometimes followed the odd instruction more literally, and I preferred that result in two of the five clips. Generation speed and clip length options felt about the same to me, no dramatic gap either way. So my working rule now is simple: dialogue goes to 3.1, weird art-direction experiments get one attempt on version 3 first. The free Veo 3.1 AI video generator comparison page I kept open laid out the feature rows without sales talk, then my own side-by-side renders settled every tie the table left behind.