Google's Veo 3.1 goes vertical with 4K and better character consistency

Anyone who has tried making an AI short film knows the two biggest headaches: your main character's face melts the second the camera changes angle, and your 16:9 widescreen clip looks terribly cropped once cut down for a phone screen.

Google rolled out a notable update to its Veo 3.1 video model to tackle both problems head-on. The core improvements center on its Ingredients to Video feature, which now supports native 9:16 vertical outputs and upscaling up to 1080p and 4K resolution.

Characters that don't melt across cuts

Up until now, maintaining visual consistency across multiple cuts has been a frustrating game of prompt roulette. Switch a character from a café to a street corner, and suddenly their clothing, hairstyle, and jawline change completely.

Veo 3.1 significantly tightens this control. By feeding the model reference ingredient images—such as a character photo, a specific setting, or an object—the system preserves identity across different scenes and camera angles. You can keep clothing and facial features identical while reusing textures and background styles.

Google also made the underlying model more expressive. Even with concise prompts, it can generate richer motion and narrative context rather than flat, static loops.

Native 9:16 meets 4K across the ecosystem

Instead of forcing mobile creators to crop horizontal videos and lose key framing, Ingredients to Video now outputs in native 9:16 portrait mode. For YouTube Shorts and other vertical feeds, that means instant, full-screen readiness.

On the fidelity side, state-of-the-art upscaling brings outputs to crisp 1080p for seamless timeline editing, or 4K for high-end production workflows. Google is rolling these updates out immediately: consumer creators can find them in the Gemini app, YouTube Shorts, and YouTube Create, while enterprise workflows get access via Flow, Google Vids, the Gemini API, and Vertex AI. All clips come tagged with Google's imperceptible SynthID digital watermark.

Why it matters

For solo creators and small video studios, this makes AI storytelling far more viable. You no longer need heavy external editing to fix distorted faces or reframe wide shots into vertical clips. Being able to draft a continuous storyline using a few reference photos directly inside Gemini or YouTube Create removes substantial production friction.

For the broader AI video landscape, Google is leaning heavily into its mobile and distribution advantages. While competitors continue chasing cinematic text-to-video trailers, Google is tailoring Veo directly for the platforms where people actually watch short videos every day.

We think shifting focus toward multi-image reference inputs is the right move for video generation. Directing characters via visual anchors mirrors how real production works, making the model feel less like a random slot machine and more like a dependable production tool.