Veo 3.1 Lets You Control Vertical Video and End Frames

Google’s Veo 3.1 sharpens control and goes vertical

Google’s newest update to its generative video stack, Veo 3.1, lands with features aimed squarely at real workflows, not just demo reels. Alongside multi-shot and consistency improvements, creators get two practical unlocks: native vertical (9:16) output and frame-accurate story control with first and last frame guidance. It is a clear escalation in the Veo vs. Sora race, and importantly, these capabilities are exposed through the Gemini API for automation-minded teams.

COEY take: This release turns Veo from “looks great” into “ships reliably.” Vertical output plus last-frame control are the kind of unsexy, high-leverage features that make short-form campaigns scale.

Vertical video, finally built in

Social-first teams no longer need prompt hacks or post processing to get portrait clips. Veo now supports native 9:16 generation through Gemini. That is a direct fit for TikTok, Reels, and Shorts, and it measurably reduces manual edits and re-renders. Early coverage also points to pricing drops across Veo tiers and a path to 1080p in certain aspect ratios, with vertical 9:16 commonly capped at 720p in many configurations.

Why this matters: most brands live and die by short form. Being able to set 9:16 up front means fewer artifacts, tighter brand framing, and content that actually fills the feed the way it should. For automation, the headline is simpler: one config, one render, fewer fixes.

Last-frame control (“End Frame”) changes how ads land

Veo 3.1 introduces first and last frame guidance, letting you steer exactly how a shot begins and ends, which is where ad creative lives or dies. End cards, offer reveals, punchlines, product lockups can now be directed to land precisely where it counts. This pairs neatly with the new scene extension capability for multi-clip continuity.

• Brand consistency: Lock in colors, typography, and CTA placement at the exact final moment.

• Sequenced funnels: Align endings across variants for clean retargeting paths.

• Lower edit debt: Fewer re-exports just to nail the last second.

Scene extension and steadier character and shot control

Veo 3.1 also emphasizes multi-shot cohesion: extending scenes without jump cuts, improving character consistency, and smoothing temporal jitter. These are not headline features, but they are the difference between “cool model” and “useful tool.” When you are automating dozens of variants, time saved on stitching and patching is margin.

Automation lens: what plugs in today vs. what is next

What is ready now

• Programmatic 9:16 outputs: Aspect ratio can be set in Gemini API requests, enabling batch generation for Shorts, Reels, and TikTok variants.

• First and last frame configs: End frame control is available via API, letting you standardize CTA landings across thousands of assets.

• Batch and templating patterns: Prompt templating plus spreadsheet or CMS feeds to Gemini API to storage or CDN to scheduler. Even without a native connector, this is straightforward via SDKs or HTTP calls.

What still needs work

• Vertical resolution ceilings: Many setups keep 9:16 capped at 720p. 1080p vertical would help premium placements.

• Event and webhook support: More robust job status callbacks and retry semantics would reduce orchestration glue code.

• Audio stems and metadata: Exposing separate tracks and licensing flags would supercharge automated QC and downstream edits.

Sora 2 vs. Veo 3.1: real options at the top

OpenAI’s Sora 2 has been pushing hard on unified audio plus video and feed-native creation, with a growing ecosystem of surfaces and safety rails. It supports vertical formats and is angling at the same mobile-first creative space. The gap is narrowing, and for teams betting on automation, that is the best news.

How we would call it right now:

• Control: Veo’s last-frame guidance and scene extension are meaningful for performance creatives. Sora’s strength is the one render that feels finished with A and V in sync.

• Formats: Both play well with vertical mobile. If you need frame-accurate CTA control, Veo 3.1 gets the nod.

• APIs and workflows: Veo’s Gemini routes are clear for production. Sora continues to evolve product surfaces, so watch its developer story if you are building conveyor belts, not sandboxes.

For the Sora side of that split, see OpenAI's Sora 2 App Debuts, Audio/Video Unified.

Multi-format reality: text, photo, video, audio

• Text to Video: Script to shot is table stakes. Veo’s improvements reduce the odds of a great prompt ending on a weird final frame.

• Image to Video: This is where marketers get leverage from moodboards and product shots. With vertical support, social-ready variants scale faster.

• Audio: Veo’s integrated audio pipeline continues to matter for ship-it-now clips, even if detailed sound design still benefits from a DAW.

• Cross-modal remixing: Expect tighter loops between static creative like logos, end cards, and product renders and motion variants in localized campaigns.

The pragmatic marketer’s view: what it changes this quarter

• Short-form paid and organic: More reliable 9:16 outputs, fewer manual fixes. That is especially helpful for always-on sprints.

• Performance creative at scale: End-frame control means consistent endings across headlines, offers, and geos. Small details lift CTR and QA.

• Localization pipelines: Programmatically swap end cards, currency, VO language, and CTAs while keeping story structure intact.

Where automation flows next

Once audio stems and more granular metadata arrive, we will see full conveyor belts: generate, auto-QC on content policy and brand color deltas, auto-mix VO and music based on locale, schedule, monitor, and retrain prompts and templates with performance data. Until then, human in the loop remains smart, but the loop is getting tighter.

Availability, pricing, and access

You can start with Veo models in Google AI Studio and move to the Gemini API when you are ready to scale. Reporting indicates lower per-second pricing and expanded aspect ratio options as part of Google’s push to make Veo more production friendly.

Summary: what is new, what it unlocks

Bottom line

Veo 3.1 is not a flashy demo drop. It is an operations upgrade. Native vertical output and last-frame control are exactly the knobs performance teams need to scale human creativity with machine precision. In the Sora 2 era, that makes Veo a real choice at the top of the stack.

Use AI Studio to test, then template your prompts, lock your end frames, and wire it into your content system. Keep humans on review and let the machines do the repetition. That is how you ship more stories, in more formats, without burning out the team.