Google video model · November 5, 2026
Google's newest Omni Flash video model. Start from a prompt, a single photo, or reference photos with short clips, and get up to 15 seconds with sound and dialogue built in, from 360p drafts to 4K. Clips start at 17 credits.
A science explainer, a drone flight, a crash zoom-out and a product shot, each with the sound the model generated alongside the picture.
Compared with 1.1, version 1.2 runs longer, holds closer to your references, delivers lines more naturally and costs fewer credits per second.
1.1 stopped at 10 seconds; 1.2 runs to 15, set in one-second steps. That is room for a full social ad, a product explainer or a scene with several beats in a single clip, so you no longer stitch two generations together. Give each beat a time range in the prompt and the camera can carry one continuous take through all of them.
1.2 keeps faces, garments, packaging type and brand colors closer to the images you give it, from the first frame to the last. In this side-by-side the left panel is the reference poster, the middle is 1.1 and the right is 1.2, both building the same vertical poster layer by layer.
Put the line in quotes and say who speaks it. 1.2 delivers dialogue more naturally, with clearer lip-sync, and keeps several speakers apart in one scene. Here a host built from one reference photo lifts a vase from a second photo, turns to the lens and says her line.
Both versions take text, a first frame or references with short clips, generate sound with every clip and offer the same resolutions and framings. They differ in length, the last frame, reference limits and price.
| Specification | Gemini Omni 1.2 Flash | Gemini Omni 1.1 Flash |
|---|---|---|
| Released | November 5, 2026 | August 27, 2026 |
| Length | 3–15 s | 3–10 s |
| Reference images | Up to 6 | Up to 10 |
| Last frame | No | Yes |
| Credits, 10 s at 720p | 164 credits | 200 credits |
| Credits, 15 s at 720p | 246 credits | Stops at 10 s |
| Best for | Most new clips, up to 15 s | Landing on a set last frame, or more than 6 reference images |
1.1 and 1.2 share these abilities, so moving a project to 1.2 doesn't change how you work.
Sound effects, ambience and speech are generated together with the picture, so there is no separate audio pass and no switch to remember. Describe what should be heard in the prompt to steer the mix.
Test framing, timing and lines at 360p, then render the version you like at a higher tier, up to 4K; the top tiers are upscaled from the model's render. Each render is a new generation, so details can change between draft and final.
Upload one image as the opening frame and describe what moves. The video starts on your shot, which suits product reveals, portraits and stills that need motion.
Give it character, product or style photos and a few seconds of footage for motion or camera, then say in the prompt which file plays which part. Faces, outfits and products stay recognizable across angles and cuts, which is what virtual try-ons and catalog clips need.
All four run in the same Workbench on Oeneo AI, so you can switch models without leaving the page. Credits are for a 8-second 720p clip with sound, a length all four offer.
| Specification | Gemini Omni 1.2 Flash | Veo 3.1 | Seedance 2.5 | Kling 4.0 |
|---|---|---|---|---|
| Developer | ByteDance | Kling | ||
| Length | 3–15 s | 4, 6, 8 s | 4–30 s | 3–30 s |
| Resolution | 360p–4K | 720p–4K | 480p–1080p | 720p–1080p |
| Aspect ratios | 16:9, 9:16 | 16:9, 9:16 | 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 | 16:9, 9:16, 1:1, 21:9 |
| Sound | Always on | On or off | On or off | Always on |
| Last frame | No | No | Yes | Yes |
| Reference inputs | 6 images, 3 clips | No | 30 images, 10 clips, 10 audio | 10 images, 5 clips |
| Edit a video you have | No | No | No | Yes |
| Credits, 8 s at 720p | 132 credits | 640 creditsVeo 3.1 Fast: 240 | 757 credits | 252 credits |
| Best for | Low-cost clips up to 15 s | Short shots with a negative prompt | Long clips with many references | Editing a clip you already have |
Copy a prompt, paste it into the Workbench above under the matching task, then swap in your own subject and setting. Gemini Omni follows plain descriptions; what makes the difference is naming the camera move, the sound and any spoken line.
Text to Video
Prompt length 450
A single continuous 15-second shot on a rainy neon street at night. 0–5 s: close-up of a street musician tuning a red electric guitar under a shop awning, raindrops bouncing off his amp. 5–10 s: the camera slowly pulls back to reveal a small crowd gathering under umbrellas. 10–15 s: he strikes the first chord, the crowd cheers, and the camera rises above the umbrellas. Sound: rain, the guitar, the crowd. No music bed, no on-screen text, no logos.Name the camera move
Push in, pull back, orbit, rise: one clear verb per shot keeps the take steady.
Spell out the sound
List the effects, the ambience and who speaks; anything you leave out is the model's guess.
Say what must stay blank
Write no logos, no text and no brand marks when a product or outfit has to stay plain.
Counting on an end frame
On Oeneo AI, 1.2 starts from one image. Describe the final shot in the prompt, or switch to Gemini Omni 1.1 Flash when the clip has to end on a picture you supply.
15 seconds from references
Reference to video stops at 10 s. For a longer take, start from text or a single photo.
Lines without a speaker
When two people talk, a quote with no name in front of it can come out of the wrong mouth.
Left open
A runner laces up bright orange sneakers on a rooftop at sunrise, then sprints toward the camera.
Closed off
A runner laces up plain bright orange running shoes with no logos, no stripes and no brand marks on a rooftop at sunrise, then sprints toward the camera.
Details you leave open get filled in, sometimes with stripes or marks that look like a real brand's. Naming what has to stay plain keeps product and clothing shots clean.
Values for Gemini Omni 1.2 Flash. The length, reference limits and prices of Gemini Omni 1.1 Flash are in the version table above.
Take a finished clip further: make it longer, sharpen it, move a character with its motion, or try the same idea on other models.
Generate a clip from a prompt, a still image, a reference or existing footage.
Open toolGemini Omni 1.2 Flash is a video generation model from Google, released on November 5, 2026 as the successor to Gemini Omni 1.1 Flash. It turns a text prompt, a single photo, or a set of reference images and short clips into a video of 3 to 15 seconds, with sound generated alongside the picture. On Oeneo AI it runs in the Workbench at the top of this page.
No. Each generation costs credits, and the price depends on the resolution and length you pick; the Generate button shows the exact amount before you start. New accounts get 10 credits when they sign up.
Choose text to video, image to video or reference to video in the Workbench above, write your prompt or upload your photo and references, set the length, resolution and aspect ratio, then click Generate. You don't need an API key or a Google account, and finished clips are saved to your creations.
Not on Oeneo AI: here Gemini Omni only makes new clips, and multi-turn editing isn't offered. To change an existing clip, switch the Workbench to video to video and pick a model that edits, such as Kling 4.0. To make a clip longer or sharper, use the Video Extender or the Video Upscaler.
No. Gemini Omni writes the soundtrack itself, so describe the sounds, the kind of voice and the exact lines in your prompt instead.
Google embeds an invisible SynthID watermark in every Gemini Omni video so it can be identified as AI-generated. It doesn't change how the video looks, and Oeneo AI doesn't remove it.
Google's safety filters block some prompts and reference images before a video is made. Reword the description or try a different image, then generate again.
Google says fast sports action and complex physics are still being improved, so scenes like a sprint finish or objects colliding may take a few tries. You can also run the same prompt on another model, such as Kling 4.0 or Seedance 2.5, from the same Workbench.
Describe the scene, add a frame or references if you have them, and get a clip with its own sound.