Google is deprecating Veo 2 and Veo 3 on the Gemini API — here's what you need to do before June 30, 2026
Veo 2 and Veo 3 model IDs on the Gemini API are being shut down on June 30, 2026. Migrate to Veo 3.1 Preview or the Enterprise GA models now.
Update, 30 June 2026: Veo 2 and Veo 3 Gemini API model IDs shut down today — last-call migration to Veo 3.1
The deadline covered in the original post is no longer a future date. As of today, veo-2.0-generate-001, veo-3.0-generate-001, and veo-3.0-fast-generate-001 are offline on both the Gemini API and Vertex AI. Any code still calling these model IDs is returning errors right now.
The Gemini API deprecations tracker confirms the shutdown is live. If you have not yet migrated, the supported paths are:
veo-3.1-generate-previewandveo-3.1-fast-generate-previewfor Gemini API usersveo-3.1-generate-001andveo-3.1-fast-generate-001via the Gemini Enterprise Agent Platform for production workloadsveo-3.1-lite-generate-previewas a new lower-cost option suited for high-volume applications and rapid iteration
Veo 3.1 also adds capabilities not present in the old models, including portrait video (9:16), video extension, first and last frame generation, and image-based direction using up to three reference images. Output supports 720p, 1080p, and 4K at eight seconds with natively generated audio.
The original migration guidance in this post still applies. The only change is that the urgency is no longer theoretical.
If your application calls any of Google’s Veo video generation models through the Gemini API, you have a firm deadline to work toward. Google has officially deprecated veo-3.0-generate-001, veo-3.0-fast-generate-001, and veo-2.0-generate-001, with a hard shutdown scheduled for June 30, 2026. After that date, calls to these model IDs will fail.
Which model IDs are affected
Three models are being retired:
| Model | Deprecated ID |
|---|---|
| Veo 3 | veo-3.0-generate-001 |
| Veo 3 Fast | veo-3.0-fast-generate-001 |
| Veo 2 | veo-2.0-generate-001 |
If any of those strings appear in your codebase, you need to update them before the end of June 2026.
Where to migrate
Google’s official replacements are the Veo 3.1 family. Depending on your setup and budget, you have a few options:
| Replacement Model | ID | Platform |
|---|---|---|
| Veo 3.1 Preview | veo-3.1-generate-preview | Gemini API |
| Veo 3.1 Fast Preview | veo-3.1-fast-generate-preview | Gemini API |
| Veo 3.1 Lite Preview | veo-3.1-lite-generate-preview | Gemini API |
| Veo 3.1 GA | GA endpoints | Gemini Enterprise Agent Platform |
The preview model IDs are available now through the Gemini API in Google AI Studio and Vertex AI. The GA versions live on the Gemini Enterprise Agent Platform (formerly Vertex AI), which is the right choice for production workloads that need SLA-backed reliability rather than preview terms.
For most developers, the straightforward swap is from veo-3.0-generate-001 to veo-3.1-generate-preview, or from veo-3.0-fast-generate-001 to veo-3.1-fast-generate-preview. The migration guide and full deprecation table are maintained at ai.google.dev/gemini-api/docs/deprecations.
What Veo 3.1 actually improves
Veo 3.1 is not a ground-up rebuild. It is a targeted refinement of the Veo 3 architecture, and the improvements are practical ones:
- Temporal consistency is better, meaning less flickering and morphing between frames.
- Text rendering inside video has improved, which matters if you generate videos with titles or captions baked in.
- Character consistency across a scene is more reliable.
- Camera controls are more precise for directing movement and angles.
- Longer clips are supported while maintaining output quality.
Veo 3.1 also retains what made Veo 3 notable in the first place: it generates video with synchronized audio, including dialogue, ambient sound, and music, all in a single pass. You do not need a separate audio layer. The model outputs at 720p, 1080p, or 4K, with AI-based upscaling available for added detail in textures, skin, and fabric.
Veo 3.1 has been in paid preview on the Gemini API since October 2025, so it is not new. The deprecation of the 3.0 models is Google completing the transition that was already underway.
The new budget option: Veo 3.1 Lite
If you are building something high-volume or cost-sensitive, there is now a third option beyond the standard and Fast variants. Veo 3.1 Lite (veo-3.1-lite-generate-preview), released on March 31, 2026, is positioned as the most cost-efficient model in the Veo 3.1 family. It is designed for rapid iteration and applications where you are generating a lot of video and cost per clip matters. Quality is lower than the full 3.1 model, but it is a meaningful addition for developers who previously found Veo pricing prohibitive at scale.
How Google handles these deprecations
Google’s deprecation process follows a consistent pattern. Announcements appear in the Gemini API changelog, with a listed shutdown date that represents the earliest the model will be retired. Google commits to advance notice before the actual cutoff, so the June 30, 2026 date is the floor, not necessarily the exact day.
That said, Google’s own policy for preview models only requires two weeks of notice before retirement. The fact that these models are getting a multi-month runway makes June 30 a generous deadline. Treat it as firm and plan accordingly.
This follows a pattern visible across the Gemini API. Gemini 2.0 Flash-Lite was shut down on June 1, 2026. Imagen 4 models (imagen-4.0-generate-001, imagen-4.0-ultra-generate-001, imagen-4.0-fast-generate-001) are being shut down on August 17, 2026. Google is actively cycling through its first-generation model IDs across the board.
What you should do now
The action is straightforward:
- Search your codebase for any of the three deprecated model ID strings.
- Update them to the appropriate Veo 3.1 model ID based on your use case and budget.
- Test your outputs before the deadline. Veo 3.1 is generally better, but any model change can affect outputs in ways worth validating against your specific prompts and workflows.
- If you are running production video generation at scale, evaluate the GA endpoints on the Gemini Enterprise Agent Platform rather than staying on preview model IDs long-term.
The full deprecation table and updated model IDs are at ai.google.dev/gemini-api/docs/deprecations. The Veo 3.1 model documentation on the Enterprise Agent Platform is at docs.cloud.google.com/gemini-enterprise-agent-platform/models/veo/3-1-generate.
The migration itself is a small code change. The risk is in not making it before the deadline.