By CompareBestAI Editorial Team
Quick Answer Google Lumiere is a research text-to-video model that Google never released to the public. Google Research introduced it in a January 2024 paper, showing short clips made with a new "Space-Time U-Net" design, but there's no app, API, waitlist or price. To make AI video with Google's technology today, use Veo 3.1, Google's current video model, released in October 2025. You can use it in the Gemini app, Google Flow, Google Vids and the Gemini API.
Google Lumiere drew wide attention in early 2024 for smooth, realistic motion, and it still shows up in searches and "AI video tool" lists. But Lumiere was a research demonstration, not a product. This guide explains what Lumiere is, how it worked, why you can't use it, how it compares with Veo 3.1, and which Google video tools you can actually use in 2026.
Lumiere vs Veo 3.1 at a glance
| Google Lumiere | Veo 3.1 | |
|---|---|---|
| What it is | Research text-to-video model | Production video generation model |
| Made by | Google Research | Google DeepMind |
| Announced | January 2024 (research paper) | October 15, 2025 |
| Public access | None | Gemini app, Flow, Google Vids, Gemini API, Vertex AI |
| Price | Not for sale | Free limits in some apps; paid through Google AI plans, Workspace and API usage |
| Native audio | No | Yes, dialogue and sound effects |
| Main capabilities | Text-to-video, image-to-video, stylized video, cinemagraphs, inpainting | Text-to-video, image-to-video, reference images, scene extension, first and last frame |
| Who it's for | AI researchers | Creators, marketers, businesses and developers |
Sources: Lumiere project page and paper (Google Research, January 2024); Google's Veo 3.1 announcement on the Google Developers Blog (October 15, 2025).
Key Takeaways
- Lumiere was never released. It exists as a research paper and demo videos, with no way for the public to use it.
- Any site selling "Lumiere AI" plans isn't selling Google's model. Google has never offered Lumiere commercially.
- Veo 3.1 is Google's current video model, with native audio and more creative control than Lumiere showed.
- The easiest way to try Veo 3.1 is Google Vids or the Gemini app, both available with a Google account.
- Lumiere's ideas still matter. Its single-pass approach to motion influenced how the field thinks about temporal consistency.
What Is Google Lumiere?
Lumiere is a text-to-video diffusion model developed by Google Research and described in the paper "Lumiere: A Space-Time Diffusion Model for Video Generation" (arXiv, January 2024). It generated short video clips, about five seconds long in the published examples, from text prompts or images.
Its key idea was the Space-Time U-Net architecture. Most video models at the time generated a few distant keyframes and then filled in the frames between them, which often produced jittery or inconsistent motion. Lumiere instead generated the whole clip in a single pass, processing it at multiple scales of space and time, which made motion smoother and more coherent.
The research demonstrated several capabilities:
- Text-to-video: generating a clip from a written prompt
- Image-to-video: animating a still image
- Stylized generation: creating video in the style of a reference image
- Cinemagraphs: animating only part of an image
- Video inpainting: filling in or replacing a region of a video
CBAI tracks Lumiere on its Lumiere AI tool page.
Can You Use Google Lumiere?No. As of October 2026, Google has not released Lumiere as an app, API or downloadable model, and has announced no plans to. When it was unveiled, coverage noted that Google did not plan a public release, partly because of the legal questions around AI-generated video.
Be careful with sites that claim to offer Lumiere. Some websites use the Lumiere name to sell subscriptions or credits. Those services run other models; they don't give you access to Google's research model.
What Replaced Lumiere? Google's Veo Models
Google's public video generation work moved to the Veo family from Google DeepMind. Veo was first shown in 2024, and Veo 3.1, released on October 15, 2025, is the current version available across Google's products.
Veo 3.1 adds features Lumiere never had:
- Native audio: dialogue, music and synchronized sound effects generated with the video
- Reference images: up to three images to keep a character or scene consistent
- Scene extension: connecting new clips to previous ones to build longer videos
- First and last frame: generating the motion between a starting and an ending image
For a full breakdown, read our Veo 3.1 review and Veo 3.1 features guide, or see the Veo 3.1 tool page.
Lumiere vs Veo 3.1: Key Differences
Availability
Lumiere is research only; Veo 3.1 is a product. This is the difference that matters most for anyone looking for a tool.
Motion and consistency
Lumiere's single-pass design was built for smooth, coherent motion in short clips. Veo 3.1 builds on years of later research and adds reference images and scene extension, giving users direct control over consistency across shots.
Sound
Lumiere generated silent video. Veo 3.1 generates video with native audio, including speech and sound effects.
Control
Lumiere's demos showed stylization, cinemagraphs and inpainting. Veo 3.1 offers practical controls in released apps: first and last frame, reference images, extension and, in Google Vids, prompt-based editing with Gemini Omni.
How to Use Google's AI Video Tools Today
| Where | What you get | Cost |
|---|---|---|
| Google Vids | Veo 3.1 clips inside a full video editor, plus AI avatars | Free with a Google account (6 AI avatar or clip generations a month); more on Google AI and Workspace plans |
| Gemini app | Generate short videos from a prompt | Included with paid Google AI plans |
| Google Flow | Google's AI filmmaking tool built around Veo | Credits included with Google AI plans |
| Gemini API and Vertex AI | Programmatic Veo 3.1 access for developers | Billed per second of generated video |
Plan limits change often; see our Veo 3.1 pricing and Google Vids features guides for current details.
Easiest starting point: open Google Vids at vids.new with a free Google account and generate a Veo 3.1 clip from a text prompt.
Veo 3.1 Alternatives
If you want options outside Google, these video generators are available today:
- Runway: creative control and VFX tools; see Veo 3.1 vs Runway
- Sora 2: OpenAI's video model; see Veo 3.1 vs Sora 2
- Kling 3.0: realistic human motion
For the full list, see Veo 3.1 alternatives or our AI video generator comparison.
How We Researched This GuideCompareBestAI follows the process described in how we rank tools and our editorial guidelines. For this guide we:
- Reviewed the Lumiere project page and research paper from Google Research
- Checked coverage of Lumiere's January 2024 announcement for release plans
- Checked Google's Veo 3.1 announcement for features and availability
- Used CBAI's Google Vids research for free-tier limits
- Removed the invented plans, prices, "googlews" comparisons and statistics from the earlier version of this page
Lumiere can't be tested because it isn't publicly available. Our comparison is based on Google's published research and announcements.
Google Lumiere FAQs
Is Google Lumiere available to the public?
No. Google Lumiere is a research model from Google Research, described in a January 2024 paper. Google has not released it as an app, API or downloadable model, and no official waitlist or pricing exists. Google's publicly available video model is Veo 3.1.
How do I use Google Lumiere?
You can't use Google Lumiere, because it was never released. To create AI video with Google's technology, use Veo 3.1 through Google Vids, the Gemini app, Google Flow or the Gemini API. Google Vids is the easiest free option with a Google account.
What is the difference between Lumiere and Veo?
Lumiere is a 2024 research model that demonstrated smooth motion with a Space-Time U-Net design but was never released. Veo is Google DeepMind's production video model family. Veo 3.1, released in October 2025, adds native audio, reference images, scene extension and wide availability across Google products.
Is Lumiere AI free?
Google's Lumiere isn't available at any price, free or paid. Websites selling "Lumiere AI" plans are not offering Google's research model. For free AI video from Google, Google Vids gives personal Google accounts 6 AI avatar or video clip generations a month using Veo 3.1.
What is a Space-Time U-Net?
A Space-Time U-Net is the architecture behind Lumiere. Instead of generating keyframes and filling the gaps, it generates a video's full duration in one pass by downsampling and upsampling the clip in both space and time. This approach aimed to make motion smoother and more consistent.
What is the best Google AI video generator in 2026?
Veo 3.1 is Google's best available video model in 2026. Use it in Google Vids for complete videos with avatars and narration, in Google Flow for AI filmmaking, in the Gemini app for quick clips, or through the Gemini API for developer projects.
Final Verdict: Google Lumiere vs Veo 3.1
Google Lumiere is an important piece of AI video research, but it isn't a tool you can use, and it never was. If you came here looking for Lumiere, the practical answer is Veo 3.1: Google's current video model, with native audio and creative controls, available today through Google Vids, the Gemini app, Google Flow and the Gemini API.
Read next: Veo 3.1 review, or browse every tool in our Video Generation category. For more Google AI tools, see 7 hidden AI tools from Google you're probably not using yet.
About CompareBestAI: CompareBestAI compares AI tools and the software businesses run on, with transparent pricing checks and a consistent scoring method. Get new AI video guides in our newsletter →
