Google Veo 3.1: The AI video generator that changes everything
Google quietly shipped Veo 3.1 and the AI video gap closed overnight. The output isn't "pretty good for AI" anymore — it's footage you'd actually pay a production crew for. Synced audio, scene control, in-video editing, dialogue. The whole stack.
I ran the full demo through Gemini, generated some of the wildest test clips I've made all year (a car explosion sequence is in the video), and broke down what it really costs in practice.
Feature 1: Realistic synced audio
The audio isn't an afterthought stitched on later — it's generated with the video. Sound effects match the visuals frame-by-frame. Footsteps land on the right step. Glass shatters at the moment of impact. This is the single biggest leap.
Feature 2: First and last frame control
Drop in the frame you want to start from and the frame you want to end on. Veo fills the motion between them. This is how you finally get scene continuity across an 8-second clip — and the foundation for stitching multi-shot sequences.
Feature 3: Combine any elements into one scene
Upload a product, a character, a setting — Veo composites them into a single coherent shot. This is the unlock for product ads, fashion content, and brand storytelling. No green screen, no compositor, no After Effects skills.
Sound effects and dialogue
The dialogue generation is what surprised me most. You can prompt actual lines and the lip sync holds. It's not perfect for a 60-second monologue yet, but for 6-second hooks and ad spots, it's production-ready.
The 8-second limit (and how to beat it)
Every Veo clip is capped at 8 seconds. The workaround is chaining — use the last frame of clip 1 as the first frame of clip 2. With first/last frame control you can build 24, 32, 48-second sequences that feel continuous.
Feature 4: In-video object removal
Spot something distracting in the frame? Brush it out. The model fills in plausible background. This was a separate $200/month tool six months ago. Now it's a checkbox.
How to test Veo in Gemini
The free path: gemini.google.com → switch to the Video model → prompt. You get a limited number of generations on the free tier. Enough to feel out the model before committing.
Mind-blowing examples
The car explosion sequence in the video is the demo I keep showing people. One prompt, 8 seconds, indistinguishable from a Hollywood VFX shot. Five years ago this was a $50k production. Today it's $0.50 in compute.
Pricing (the part Google buries)
Veo on Gemini's paid tier sits at the $20/month Pro plan for limited use, with the $200+/month Ultra tier for serious volume. If you're generating daily, the higher tier pays for itself fast.
The all-in-one alternative
If you want Veo-class video generation plus scripts, voiceover, music, thumbnails, and editing all in one login — the AI Media Machine bundles the whole stack for $1 trial. Same outputs, fraction of the cost, no juggling subscriptions.
