The endpoints that made Veo 3.1 callable from code — Standard, Fast, and Lite — are all listed for retirement no sooner than October 22, with Google pointing Gemini API developers to gemini-omni-1.1-flash as the replacement.
2026-10-18 · by Moe Ameen
Google's Gemini API deprecations page lists all three Veo 3.1 preview models — `veo-3.1-generate-preview` (Standard), `veo-3.1-fast-generate-preview` (Fast), and `veo-3.1-lite-generate-preview` (Lite) — with an earliest possible shutdown date of October 22, 2026, and names `gemini-omni-1.1-flash` as the recommended replacement for all three. Google is explicit that the listed dates are the earliest a model might be retired and that the exact cutoff will be communicated in advance, so October 22 is a floor, not a confirmed switch-off. Anyone calling a `veo-3.1-*-preview` ID from code is on notice to migrate.
The [Veo 3.1 API](/ai-tools/veo-3-1-api) is the programmatic way to generate short video with synchronized native audio from a prompt, as opposed to clicking through the Gemini app or Google Flow. Standard and Fast launched on the Gemini API on October 15, 2025; the cheaper Lite tier followed on March 31, 2026, and a January 2026 update added the "Ingredients to Video" reference-image workflow, native 9:16 portrait output, and 1080p/4K resolutions. All three shipped as preview models from day one, which is the status that is now expiring — preview endpoints were never a long-term contract.
Google's newer video work has been surfacing under the Gemini Omni name rather than as a Veo release, and `gemini-omni-1.1-flash` is where the Gemini API now routes developers. Early reports from developers migrating suggest Omni's per-clip billing differs from the Veo tiers, so re-check your cost model rather than assuming a like-for-like swap; confirm current pricing on Google's own pages. Teams that specifically need Veo can look to the Vertex AI production models (the `-001` IDs), which sit on a separate, later retirement timeline aimed at Google Cloud deployments. The churn is not unique to Google — OpenAI shut down its Sora 2 video API in September 2026 without a direct replacement — so model IDs across the whole category are proving short-lived.
As always with deprecations, the primary source is the one that matters: check Google's Gemini API deprecations page for the final date and the current replacement before you change any code, since preview-model timelines can move.
This deprecation is a clean illustration of why you should not bolt your content operation directly to one generation model's API. The clip generator is the volatile part — preview IDs retire, model families get renamed, per-second pricing shifts, and a workflow welded to `veo-3.1-generate-preview` becomes a rewrite the day Google flips the switch. [Kompozy](/) sits one layer up, and it is deliberately model-agnostic: it ingests the finished MP4, not the endpoint that made it. Whether that file came from Veo 3.1 today, `gemini-omni-1.1-flash` next week, or the Vertex AI `-001` models after that, the downstream job is identical.
So the move while this migration is in flight is to keep producing and let the finishing layer absorb the churn. Generate your clips through whichever Google endpoint survives, drop the exports into Kompozy, and [Clipped Shorts](/glossary/clipped-short) cuts each one to vertical 9:16 with word-synced captions for muted autoplay feeds, while [HyperFrames](/glossary/hyperframes) wraps a brand-exact carousel, quote graphics, and photo posts around the same idea — all held to one voice by your [Persona Brief](/glossary/persona-brief). Between external renders, Kompozy's own face-locked [Persona Shorts](/glossary/persona-shorts) keep a consistent avatar on camera with no third-party model at all, so your series never goes dark during a deprecation window. [Autopilot](/glossary/autopilot) then schedules and publishes the whole set across the eight social platforms plus blog and email behind a per-post review. Let Google reshuffle its video endpoints; your publishing pipeline shouldn't notice.
Google's Gemini API deprecations page lists all three Veo 3.1 preview models — veo-3.1-generate-preview, veo-3.1-fast-generate-preview, and veo-3.1-lite-generate-preview — with an earliest possible shutdown of October 22, 2026. Google notes that date is the earliest a model might retire and that the exact cutoff will be announced in advance, so confirm the final date on Google's own page before relying on it.
Google points developers to gemini-omni-1.1-flash as the recommended replacement for all three Veo 3.1 preview IDs. It is part of the Gemini Omni family rather than a newer Veo, so output and per-clip billing can differ — re-test rather than assuming a drop-in swap. If you need Veo specifically, the Vertex AI production -001 models sit on a separate, later timeline.
The deprecation notice concerns the Veo 3.1 preview model IDs exposed through the Gemini API for programmatic access. App-based and Flow access to Google's video models follows Google's product decisions separately. If you call the API from code — directly or through a third-party tool — that is the surface to check and migrate.
Decouple generation from finishing and publishing. Keep the volatile part — the clip model — swappable, and run the exported files through a model-agnostic engine like Kompozy that clips, captions, reframes, builds carousels, and schedules across the eight social platforms plus blog and email. When the underlying model ID changes, your downstream pipeline is unaffected because it works from the MP4, not the endpoint.