Gemini Omni 1.2: what changes for your social media videos
Released on November 5, 2026, Gemini Omni 1.2 makes longer, cheaper clips than 1.1. What changed, what it costs, and when to pick it for a Reel or a TikTok.
By Théo, Oposto team · · Tools
Gemini Omni 1.2 Flash, released on November 5, 2026, makes clips of 3 to 15 seconds instead of 10 at most, costs about $0.07 per second at 720p (30% less than 1.1), and can edit or extend a clip it created up to 30 seconds in total. For a Reel, TikTok or Short built from one simple scene, it is now the easiest and cheapest option from Google. In Oposto, it is the model for simple scenes in the video studio and the storyboard, with length, quality and edits set scene by scene.
Up to date as of November 9, 2026. Every figure here comes from Google's documentation (the Omni page of the Gemini API, the model page, the pricing page) and the November 5 announcement. For a comparison with Seedance, Kling and the rest, see our guide to AI video for social media.
What is Gemini Omni 1.2 Flash?
It is Google's video model that creates a clip with sound from text, images or video, then edits it through conversation. Its API name is gemini-omni-1.2-flash. It replaces Gemini Omni 1.1 Flash, last updated in August 2026. Google offers it in the Gemini API and Google AI Studio, Google Cloud, Flow, Google Vids, the Gemini app (for Google AI Ultra subscribers) and Google Ads.
What matters if you post short videos:
- Longer clips: 3 to 15 seconds, set to the exact second.
- A lower price: about $0.07 per second at 720p, against about $0.10 for 1.1.
- Four quality levels: 360p, 720p, 1080p and 4K, the last two upscaled.
- Up to 6 reference images for a person, a product or a style.
- A first frame and a last frame, with the motion in between.
- Editing and extending a clip it already made, without uploading it again.
What changed between Omni 1.1 and Omni 1.2?
| Feature | Omni 1.1 Flash | Omni 1.2 Flash |
|---|---|---|
| Clip length | 3 to 10 s | 3 to 15 s, by the second |
| With first and last frame | 3 to 10 s | 3 to 10 s |
| Video you can edit | Up to 10 s | Up to 15 s |
| Extension | Input video up to 10 s | 3 to 10 s per turn, 30 s in total |
| Sound | Video only listed as output | Video with native audio |
| Quality | 360p, 720p, 1080p, 4K | 360p, 720p, 1080p, 4K |
| Price at 720p | About $0.10 per second | About $0.07 per second |
The rest stays the same: 16:9 and 9:16 formats, 24 frames per second, and no free tier in the API. Google also promises better adherence to reference images, continuous camera moves with fewer unwanted jump cuts, more natural dialogue and clearer lip sync. These are Google's claims, not public measurements, so judge them on your own tests.
One practical point: version 1.1 stays available for a four-week transition. From December 3, 2026, requests sent to Omni 1.1 are routed to Omni 1.2 automatically and billed at 1.2 rates.
Omni 1.2 Flash in four numbers
- 15 s Longest clip in one go (10 s with 1.1)
- 30 s Total length after extensions
- 6 Reference images at most
- $0.07 Per second at 720p, 30% less than 1.1
How long can a Gemini Omni 1.2 video be?
A single clip runs 3 to 15 seconds, 10 by default. You pick the exact length, to the second: 7 seconds for a product shot, 15 for someone talking to camera. Above 15 seconds the API rejects the request. With a first and a last frame, the limit drops to 10 seconds.
To go further, you extend the clip: each turn adds 3 to 10 seconds at the end, up to 30 seconds in total. The model looks at the last 10 seconds to carry the motion on. Extension only adds to the end, never to the start.
For a Reel, 15 seconds in one go is often enough. To pick the right length for each platform, read how long a Reel should be.
Which quality should you pick, and what does it cost?
The price depends on quality and length. Here are the rates from the Gemini API pricing page:
| Quality | Approx. price per second | Best for |
|---|---|---|
| 360p | $0.023 | Tests, rough drafts |
| 720p | $0.07 | Most videos watched on a phone |
| 1080p (upscaled) | $0.104 | Product shots, ads |
| 4K (upscaled) | $0.209 | Rarely needed for social media |
Google gives two 720p examples: about $0.70 for a 10 second clip and about $1.04 for 15 seconds. Its suggested workflow is simple: draft at 360p or 720p, fix details with follow-up edits, then upscale to 1080p or 4K only at the end. For a video watched on a phone, 720p is usually enough. Exact sizes per platform are in our image and video size cheat sheet.
How do reference images and first and last frames work?
You can give up to 6 reference images in total. For a person, up to 3 images each, for up to 2 people. An image can also show a product, a place or a style to follow.
The change to know about: tags such as <IMAGE_REF_0> are deprecated by Google. Describe each image's role in plain sentences instead, for example "Image 1: the customer, keep her exactly as she is. Image 2: the bottle, label facing the camera. Image 3: the colour mood to follow".
For a precise move, give two images, the first then the last, and describe what happens in between. It works well for a before and after, a product being unboxed, or a room going from empty to tidy. The length then stays between 3 and 10 seconds.
Google also supports short motion clips as references (3 seconds each at most). Their audio is ignored, and they only work when creating a new clip, not when editing or extending one.
How do you edit or extend a clip you already made?
Omni 1.2 keeps track of the conversation. Once the clip exists, you ask for a change: a different product or background, new lighting, other weather, another aspect ratio. The model reworks the clip it just made, up to 15 seconds, without you uploading it again. The rest of the shot stays put.
From first clip to 30 seconds
- Create A 3 to 15 s clip, at 360p or 720p to start.
- Edit Background, product, light: the model reworks its own clip.
- Extend 3 to 10 s more per turn, up to 30 s in total.
- Finish Upscale to 1080p or 4K once the shot is approved.
An important limit for the UK, Switzerland and the European Economic Area: editing or extending a video you upload is not available there. Editing and extending a video created by the model is. Also note that you cannot edit the voice, and audio inputs are not supported.
Who is Gemini Omni 1.2 the right choice for?
| Your video | Omni 1.2? | Why |
|---|---|---|
| One simple scene, 15 s or less | Yes | Low price, exact length, sound made with the picture |
| Someone talking to camera | Yes, or your own take | Dialogue and lip sync announced as more natural; a real take is still the most credible |
| A before and after, an unboxing | Yes | First and last frame |
| A 20 to 30 s shot without cuts | Yes, by extension | 3 to 10 s added per turn |
| An announcement with text to read | Not really | Text is drawn like a picture; motion design gives sharp text |
| The most polished shot possible | Compare first | Other models, such as Seedance 2.5, aim for a more crafted, pricier look |
In short: Omni 1.2 suits small brands and freelancers who post often, want a low cost per video and keep scenes simple. For a video people mostly need to read (numbers, a list, a promotion), go with motion design. If you would rather stay off camera, see our Reel ideas for when you don't like being on camera.
How do you write a prompt for Gemini Omni 1.2?
A good prompt states the length, the format, what we see, what we hear and the role of each image. Three examples to adapt (they are made up):
Vertical 9:16 video, 12 seconds. A baker takes a tray of croissants out of the oven while the camera slowly pushes in on the tray. We hear the oven door and the pastry crackling. Image 1: the baker, keep her exactly as she is. No text on screen.
Vertical video, 8 seconds. Image 1: the empty room, at the start. Image 2: the same room furnished, at the end. The camera stays still and the furniture appears piece by piece. Soft music.
Extend this clip by 6 seconds: the customer walks out of the shop with her bag and heads right. Keep the same light and the same background sound.
Three tips: give the length in seconds, describe the sound as carefully as the picture, and do not ask for on-screen text (add it in the edit). For the structure of a short video, see how to automate short-form video.
How is Gemini Omni 1.2 used in Oposto?
In Oposto you don't have to set the model: you describe the video in the chat or in the video studio, and Oposto suggests a model for each scene, which you can change. Omni 1.2 handles simple scenes and is available from the free plan. What version 1.2 changes for you:
- Scenes up to 15 seconds: the scene length is the clip length, with no extra cut. When someone speaks, the speech sets the length; if the script runs long, the text gets shorter, never the voice.
- Quality scene by scene: 360p, 720p, 1080p or 4K, so you only pay for high quality on the shots that need it.
- Up to 6 images per scene, each with its role written in plain words. A last frame you set on the scene becomes the final frame of the clip.
- Edit or extend a scene made by Omni, from the storyboard or by asking the storyboard chat: an extension adds 3 to 10 seconds, up to 30 seconds, and the scene grows to match.
- Studio b-roll: a shot created in the studio can run up to 15 seconds and be edited or extended the same way.
- The assistant knows all this: you can ask it to change a scene's quality, edit it or extend it.
With Omni, Oposto only edits and extends clips that Omni created, precisely because that is what Google allows in Europe. A video you film yourself goes through the edit, not through Omni. The rest stays the same: your brand DNA goes with every request, the model never draws an app screen, and captions stay clear of the areas hidden by Instagram and TikTok. Some video features (motion design, other models) need a paid plan, from the Pro plan for motion design.
Frequently asked questions
What is the difference between Gemini Omni 1.1 and 1.2?
Omni 1.2 makes clips of 3 to 15 seconds (10 at most with 1.1), edits videos up to 15 seconds, extends a clip up to 30 seconds in total and costs about 30% less at 720p.
How much does Gemini Omni 1.2 cost?
In the Gemini API, about $0.023 per second at 360p, $0.07 at 720p, $0.104 at 1080p and $0.209 at 4K. A 15 second 720p clip comes to about $1.04.
How long can a Gemini Omni 1.2 video be?
15 seconds in one go, or 10 seconds with a first and last frame. By extending the clip 3 to 10 seconds per turn, you can reach 30 seconds in total.
Is Gemini Omni 1.1 going away?
From December 3, 2026, requests sent to Omni 1.1 are routed to Omni 1.2 automatically and billed at 1.2 rates.
Can I edit videos with Gemini Omni 1.2 in the UK or EU?
Yes for videos the model created. Editing or extending a video you upload is not available in the UK, Switzerland or the European Economic Area.
Does Oposto use Gemini Omni 1.2?
Yes, for simple scenes in the video studio and the storyboard: scenes up to 15 seconds, quality chosen per scene, and editing and extension of clips Omni created.
Want to see a 15 second scene for your own brand? Describe your video to the assistant and let Oposto pick the model.