NewReleased November 5, 2026 · live on imageatGemini Omni Flash 1.2
15-second clips. Up to 4K.
Google's fast video model makes 3 to 15 second videos with native audio from text, images or references, and edits, extends and upscales your clips. Generate it in 4K on imageat, from 1 credit per second.
Gemini Omni Flash 1.2 at a glance
Gemini Omni Flash 1.2 (officially “Gemini Omni 1.2 Flash”) is Google DeepMind's fast, cost-effective AI video model, released and made generally available on November 5, 2026. It generates 3–15 second videos with native audio from text, a start and end frame, or up to 10 reference images, in 16:9 or 9:16, and edits, extends and upscales existing clips. Google prices it at $0.07 per second at 720p, 30% less than Omni 1.1. On imageat it runs at 360p, 720p, 1080p or 4K for 1, 3, 4 or 7 credits per second.
Updated November 5, 2026 · Sources: Google announcement, fal model pages
Gemini Omni 1.2 in motion
One official sample for each of the six modes, served from imageat's CDN. Every output carries Google's invisible SynthID watermark.
Text to video
A 10-second editorial film
One long prompt written as beats: a ceramic artist opening her seaside workshop at sunrise, with sound.
Image to video
A still photo comes alive
A kingfisher photo becomes a quiet, continuous wildlife shot with natural ambience.
Reference to video
Three references, one scene
A character, a product and a place from three images, combined into a night-market delivery scene.
Edit
White tulips turned crimson
"Change all white tulip petals to rich crimson red… keep the vase, background, lighting and camera movement unchanged."
Extend
Continue the scene
An existing clip gets 8 more seconds that pick up from its last frames, with matching motion and sound.
Upscale
Draft to 1080p
A low-resolution draft upscaled to 1080p. The same endpoint goes to 4K for the takes you keep.
What's new in Gemini Omni 1.2
Omni 1.2 Flash is built for video at scale: longer clips, steadier characters and products, conversational editing and a lower price per second.
15-second clips
Up to 15 seconds per generation, up from 10 in Gemini Omni 1.1. Pick any length from 3 to 15 seconds, one second at a time, so you pay for exactly the shot you need.
Closer to your references
Guide the video with character, product and style images. Omni 1.2 holds faces, logos and products steady from the first frame to the last, with less drift and morphing.
Multi-turn editing
Edit a clip by describing the change: swap the product or background, change the lighting or weather, or restyle it. Then edit the result again until it is right.
Natural dialogue
More fluid speech, clearer lip sync and distinct voices when several characters talk. Write the lines in quotes and the model performs them.
Fewer jump cuts
Sweeping FPV flythroughs, dramatic zoom-outs and orbits come out as one continuous move instead of being broken up by unwanted cuts.
Draft cheap, finish in 4K
Draft at 360p or 720p, then upscale the take you like to 1080p or 4K. On imageat you can also generate directly in 4K.
Gemini Omni Flash 1.2 modes
and input limits
Six modes, all in the imageat video studio. Every generated clip includes native audio.
| Mode | Input | Output |
|---|---|---|
| Text to video | A prompt | 3–15s · 360p–4K · 16:9 or 9:16 · audio |
| Image to video | A start image, plus an optional end image to land on | 3–15s (≤10s with an end frame) |
| Reference to video | Up to 10 images, referred to as <IMAGE_REF_0>, <IMAGE_REF_1>… | 3–15s · 360p–4K |
| Edit | A source clip up to 15s and an instruction | Keeps the source length and framing |
| Extend | Any clip and a prompt for what happens next | Adds 3–10 new seconds |
| Upscale | A source clip up to 15s | 720p, 1080p or 4K |
How to prompt Gemini Omni 1.2
Write the camera
Name the move and say it is continuous: "one unbroken FPV shot, no cuts" or "a slow orbit around the bottle".
Script the sound
Put dialogue in quotes and name who says it, then list ambience and effects: waves, gulls, a café crowd.
Point at references
Call each upload by its tag: "the jacket from <IMAGE_REF_1> on the model in <IMAGE_REF_0>".
Edit one thing at a time
Say what changes and what stays: "make the sky stormy, keep the people, framing and camera move".
Gemini Omni 1.2 vs Omni 1.1
| Capability | Gemini Omni 1.2 Flash | Gemini Omni 1.1 |
|---|---|---|
| Availability | Generally available (Gemini API, Vertex AI) | Earlier release |
| Clip length | 3–15 seconds, 1-second steps | Up to 10 seconds |
| Price (Google, 720p) | $0.07 per second | About 30% higher |
| Reference adherence | Character, product and style images, short motion clips (≤3s), audio cues | Follows references less closely |
| Consistency | Less drift and morphing from first to last frame | Baseline |
| Editing | Multi-turn: product, background, lighting, weather, aspect ratio | Baseline |
| Dialogue | Clearer lip sync, distinct multi-speaker voices | Baseline |
Based on Google's November 5, 2026 announcement. Google notes that fast sports action, rapid scene changes and complex physics are still being improved.
Gemini Omni 1.2 price per second
An 8-second 720p clip is 24 credits; the same clip in 4K is 56. Edits bill each second of the source clip, extensions only the new seconds, and upscaling costs 2, 3 or 7 credits per second to 720p, 1080p or 4K. Failed generations are refunded automatically.
Gemini Omni 1.2 API
Call it through the imageat video API or MCP server with one key for every model.
POST https://api.imageat.com/v1/videos
{
"model": "gemini-omni-12",
"prompt": "One unbroken FPV shot over a fjord...",
"duration": "8s",
"resolution": "4k",
"aspectRatio": "16:9"
}Make your first
4K Gemini Omni video
Gemini Omni Flash 1.2 sits in the imageat video studio next to Kling 4.0, Seedance 2.5, Wan 3.0 and Veo. Write the shot, pick a length and resolution, and pay only for the seconds you generate.
Gemini Omni 1.2 questions
What is Gemini Omni Flash 1.2?
+
Gemini Omni Flash 1.2 (Google calls it Gemini Omni 1.2 Flash) is Google DeepMind's fast, cost-effective video model, released on November 5, 2026. It makes 3 to 15 second videos with native audio from text, a start image, first and last frames, or character, product and style references, and it edits, extends and upscales existing clips. On imageat it runs in the AI video studio at up to 4K.
What is new in Gemini Omni 1.2 compared to 1.1?
+
Clips go up to 15 seconds, up from 10 seconds in Omni 1.1, and you can set the length in one-second steps from 3 to 15 seconds. Google cut the price by 30%, improved how closely the video follows reference images, reduced drift and morphing so characters and products stay consistent from the first to the last frame, made dialogue and lip sync more natural with distinct multi-speaker voices, and cut down on unwanted jump cuts in camera moves like FPV flythroughs, zoom-outs and orbits.
Can Gemini Omni Flash 1.2 make 4K video?
+
Yes. On imageat you can generate Gemini Omni Flash 1.2 clips at 360p, 720p, 1080p or 4K (3840×2160), and upscale an existing clip of up to 15 seconds to 1080p or 4K. Google recommends drafting at 360p or 720p and upscaling your best takes, which is the cheapest way to get a 4K final.
How long can a Gemini Omni Flash 1.2 video be?
+
3 to 15 seconds per generation, chosen one second at a time. With a first and last frame the clip is up to 10 seconds. Extend adds 3 to 10 new seconds to any clip, so you can chain extensions into a longer sequence.
Does Gemini Omni Flash 1.2 generate sound?
+
Yes. Every clip comes with synchronized audio generated with the picture: dialogue with lip sync, multiple distinct speakers, ambience and sound effects. Put spoken lines in quotes in the prompt and describe the sounds you want.
How much does Gemini Omni Flash 1.2 cost on imageat?
+
It is billed per second of video: 1 credit per second at 360p, 3 at 720p, 4 at 1080p and 7 at 4K. An 8-second 720p clip is 24 credits and an 8-second 4K clip is 56. Edits bill each second of the source clip, extensions bill only the new seconds, and upscaling is 2, 3 or 7 credits per second to 720p, 1080p or 4K. Failed generations are refunded automatically.
How do I use reference images with Gemini Omni Flash 1.2?
+
Choose Reference to Video in the AI video generator and add up to 10 images of characters, products, places or styles. Refer to them in the prompt in upload order as <IMAGE_REF_0>, <IMAGE_REF_1> and so on, for example: "The model in <IMAGE_REF_0> turns 360 degrees wearing the jacket from <IMAGE_REF_1>."
Can I edit a video with Gemini Omni Flash 1.2?
+
Yes. Upload a clip of up to 15 seconds in the video studio's edit mode and describe the change: swap a product or the background, change the lighting or weather, or restyle the whole clip. The edit keeps the source length and camera motion, and you can edit the result again for multi-turn changes.
Are Gemini Omni Flash 1.2 videos watermarked?
+
Google embeds an imperceptible SynthID watermark in every Gemini Omni output so it can be identified as AI-generated. imageat adds no visible watermark on paid plans.
Is Gemini Omni Flash 1.2 available through an API?
+
Yes. Google offers it through the Gemini API, Google AI Studio and Vertex AI. On imageat you can call it with the model id gemini-omni-12 through the imageat video API or MCP server, with one key for every image and video model.
Gemini is a trademark of Google LLC. imageat is not affiliated with Google. Read Google's announcement.