Is Gemini Omni 1.2 Flash generally available?

Yes. Google made it generally available on November 5, 2026, through the Gemini API, Google AI Studio, Google Cloud and several Google products.

How long can a generated clip be?

Up to 15 seconds, set in one-second increments from 3 to 15 seconds. That is five seconds longer than Omni 1.1 Flash.

Which resolutions does it output?

360p, 720p, 1080p and 4K. The intended workflow is to draft at 360p or 720p, edit, then upscale the chosen take.

Does it generate audio as well as video?

Yes. Dialogue, lip-sync and ambience are generated with the picture, and separate voice references let it handle multiple speakers.

What inputs and references does the model accept?

A text prompt, character or product images, a style frame, video for editing, motion clips of up to three seconds and audio samples. All of them can be combined in one request.

Can I edit a clip without generating it again?

Yes. Multi-turn conversational editing lets you swap a product or background, change lighting or weather, or switch aspect ratio on the same clip.

Can it connect two keyframes, extend a clip or upscale old video?

Yes to all three. It interpolates between a first and last frame, extends an existing clip, and upscales previously generated video as a standalone step.

Is Gemini Omni 1.2 Flash available on OnTrend AI right now?

Not yet. A dedicated studio is in development. Until it opens, OnTrend AI templates remain the way to create on this site.

How much does the model cost?

Google lists roughly seven US cents per second of 720p video, about 30% below Omni 1.1. OnTrend AI credit pricing is separate and will be published with the studio.

How is 1.2 different from 1.1 Flash?

Longer clips, stronger reference following, multi-turn editing, more natural speech and a 30% lower price. Resolution options and the SynthID watermark carry over.

Can I use the generated videos commercially?

Commercial terms come from Google's model terms and your account type, not from this page. Outputs also carry an imperceptible SynthID watermark, so check the current terms before a paid campaign.

What does the model still struggle with?

Google names fast sports action, rapid script changes and complex physics as areas still being improved. Build those shots with extra takes or a simpler fallback.

What is SynthID and does it affect my edits?

SynthID is Google's imperceptible watermark embedded in generated video. It is designed to survive edits and compression, and it does not change how the clip looks.

Where can I try it without writing code?

Google AI Studio, Flow, Google Vids and the Gemini app for AI Ultra subscribers offer no-code access. Developers and businesses use the Gemini API or Google Cloud.

NEW MODEL · GA NOV 5, 2026

Gemini Omni 1.2 Flash: 15-Second AI Video With Native Audio

Q: Does it keep characters and products consistent?

Reference adherence is the headline improvement in 1.2. Faces, garments, packaging type and brand colours are designed to hold across angles and shots with less drift.

Google's omni-modal video model turns a written scene, reference images and short motion clips into a single 15-second video with synchronized sound. Here is what the model can do, where it runs, and what OnTrend AI is building on top of it.

Explore Viral Templates See What It Can Do

✓Native 15-second clips
✓3s to 15s in one-second steps
✓Draft at 360p / 720p, finish at 4K
✓Synchronized audio

GEMINI OMNI 1.2 FLASH // 15S NATIVE // REFERENCE-GUIDED // MULTI-TURN EDITING // MULTI-SPEAKER AUDIO // SYNTHID WATERMARK // 30% LOWER PRICE

Official Gemini Omni 1.2 Flash Demo Clips

Six examples published by Google with the 1.2 Flash launch: virtual try-on, reference-led typography, multi-speaker dialogue, continuous camera moves, motion references and a 15-second product explainer.

VIRTUAL TRY-ON

One Model Photo, Three Outfits

A single model photo plus front-and-back product shots produces a 15-second try-on sequence with 360-degree turns and match cuts.

REFERENCE ADHERENCE

Reference-Led Typography

A static style frame becomes a layered 15-second title sequence while materials, shadows and letterforms stay locked.

NATIVE AUDIO

Multi-Speaker Dialogue

Four character references and one-second voice samples route each voice to the right speaker with matching lip-sync.

CAMERA CONTROL

One Continuous FPV Flight

Two architectural references guide an uninterrupted drone move through the Pantheon dome, with zero jump cuts.

MOTION REFERENCE

Motion From a Three-Second Clip

Turn one establishes an octopus; a follow-up turn casts a dancer who mirrors the same fluid motion.

15-SECOND EXPLAINER

Macro Product Explainer

A four-shot macro explainer with correctly spelled, motion-tracked callouts, generated as one 15-second clip.

Demo clips were published by Google with the Gemini Omni 1.2 Flash announcement. They are shown here for illustration only and were not generated on OnTrend AI.

What Is Gemini Omni 1.2 Flash?

Gemini Omni 1.2 Flash is Google's omni-modal video model, generally available since November 5, 2026. It accepts text, images and video, and returns video with synchronized audio up to 15 seconds long.

It sits in the fast, cost-efficient tier of Google's Omni family. Version 1.2 is an upgrade of Omni 1.1 Flash: clips grow from 10 to 15 seconds, reference following gets noticeably tighter, and the price drops 30% to roughly seven US cents per second of 720p video. Editing also moves into conversation — you refine the same clip turn by turn instead of regenerating from scratch.

What separates Omni 1.2 Flash from earlier photo-to-video tools is how much it can hold onto. A character sheet, a product photo, a style frame, a motion clip of up to three seconds and an audio sample can all act as references at once. Faces, garments, packaging typography and brand colours are meant to survive changes of camera angle and shot, with less drift between the first and last frame.

The model runs through the Gemini API and Google AI Studio for developers, the Google Cloud console for production workloads, and Google products such as Flow, Google Vids, the Gemini app for AI Ultra subscribers and Google Ads. Outputs carry an imperceptible SynthID watermark.

OnTrend AI is not running this model yet. The pages on this site are template-first: you pick a trend-tested video template, upload one or two photos and get a finished clip. The Gemini Omni 1.2 Flash studio we are building will add prompt-level control on top of that workflow, and this page tracks the capability it will bring.

How Gemini Omni 1.2 Flash Turns References Into Video

You describe the shot and attach references; the model returns one clip with picture and sound already synchronized, then keeps editing the same clip in conversation.

INPUT 01

A written scene

Describe the action, the camera move, the lighting and the dialogue. A single prompt can carry a complete 15-second story instead of a single beat.

INPUT 02

Reference images

Character sheets, product photography or a style frame anchor identity, wardrobe, packaging type and colour. Front and back product shots help 360-degree moves stay correct.

INPUT 03

Motion and audio cues

A short clip of up to three seconds sets gait, gesture or fabric movement. One-second voice samples let the model assign a different voice to each speaker.

One video, up to 15 seconds, with sound

The output is a single clip rather than a timeline of separate shots. Duration is set in one-second steps between 3 and 15 seconds, and the audio is generated with the picture rather than added afterwards.

▸Draft at 360p or 720p, then upscale the keeper to 1080p or 4K
▸Native dialogue, lip-sync, ambience and multi-speaker conditioning
▸Multi-turn edits: swap a product, background, lighting, weather or aspect ratio without a reshoot
▸First-to-last keyframe interpolation and clip extension for longer sequences
▸Imperceptible SynthID watermark on every output

This describes the model as documented by Google. OnTrend AI does not run Gemini Omni 1.2 Flash today; the studio that will is in development.

Who Gemini Omni 1.2 Flash Is Built For

Google positions 1.2 Flash for everyday video made at volume — product clips, try-ons and ads — rather than one-off cinematic experiments.

E-commerce and DTC brands

Turn a model photo and a handful of product shots into try-on sequences, catalogue spins and match-cut ads. Reference adherence keeps prints, seams and packaging text intact across angles.

Performance marketing teams

Build one master clip, then edit background, offer or aspect ratio turn by turn to produce variants for different audiences and placements. No re-brief and no separate shoot.

Creative studios and agencies

Hold a character or product consistent from the first frame to the last, interpolate between two keyframes, or extend a favourite shot into a longer sequence for a client cut.

Creators and educators

Produce explainers with motion-tracked callouts, or short dialogue scenes with distinct voices. A 15-second limit is enough for a hook, a demonstration and a call to action.

Marketplaces and retail media

Give sellers a repeatable way to turn catalogue photography into seasonal promos and sponsored placements, with product details held consistent across variants.

Local businesses and franchises

Produce storefront, service and offer spots without a crew or a shoot day. Draft at low resolution, approve the take, then finish the one that runs.

Gemini Omni 1.2 Flash Capabilities, Shot by Shot

The practical list: what you can ask for, what comes back, and the limits Google still flags.

01LENGTH

3 to 15 seconds, one-second steps

Clips grew from 10 seconds in Omni 1.1 to 15 seconds in 1.2. You choose any duration in one-second increments, which is enough for a full social spot rather than a single beat.

02RESOLUTION

Draft small, finish at 4K

Generate a draft at 360p or 720p, fix the details with follow-up edits, and upscale only the take you keep to 1080p or 4K. The tiered workflow is how Google keeps cost per finished clip down.

03REFERENCES

Character, product, style, motion, voice

Combine character or product images, a style frame, a motion clip of up to three seconds and audio cues in one request. Faces, garments, packaging type and brand colours are held across camera angles.

04EDITING

Multi-turn conversation, no restart

Swap a product, change the background, relight the scene or switch the aspect ratio by talking to the same clip. Edits work across the full 15 seconds, so ad variants do not need a fresh generation.

05AUDIO

Multi-speaker dialogue and lip-sync

Speech sounds more natural than in 1.1, lip-sync is clearer, and separate voice references let the model route lines to the correct speaker with distinct timing and delivery.

06CAMERA

Continuous moves, fewer jump cuts

Ask for an FPV flythrough, a crash zoom-out or an orbit and the model holds scale and momentum through the move, with far fewer unwanted cuts than earlier versions.

07CONSISTENCY

First frame to last, and beyond

Interpolate between a start and an end keyframe, or extend a finished clip into a longer sequence. Products, characters, props and environments are meant to stay stable from shot to shot.

08LIMITS

What still needs work

Google still flags fast sports action, rapid script changes and complex physics as areas under improvement. Treat those as the current edge of the model, not as guaranteed output.

How to Make a Video With Gemini Omni 1.2 Flash Today

The model is live in Google's developer and business tools. OnTrend AI's dedicated studio is not open yet, so the fastest route on this site is still a template.

Open the model in Google's tools

Prototype in Google AI Studio or call the model through the Gemini API. The Google Cloud console and Gemini for business expose the same model for production workloads.

Gemini API · AI Studio · Google Cloud

Attach references and describe the shot

Add character or product images, a motion clip of up to three seconds and audio cues, then write the scene, camera move and dialogue into the prompt.

Images · video ≤ 3s · audio samples

Edit in conversation, then upscale

Keep the take that works, change one detail per turn, and upscale the final clip to 1080p or 4K once the story is right.

Draft 360p / 720p → finish 1080p / 4K

Or start with a ready-made template

Reference Material That Holds Up

Omni 1.2 Flash follows references more closely than 1.1, but the input still sets the ceiling. Prepare these before you prompt.

Recommended Best Practices

✓

Front-facing character sheets

A clean, evenly lit portrait or full-body shot gives the model a stable identity anchor for every angle.

✓

Product shots from both sides

Front and back photos keep packaging type, seams and prints correct through 360-degree turns.

✓

One clear motion clip

A single-subject clip of up to three seconds is enough to carry gait, gesture or fabric movement.

✓

One voice sample per speaker

A one-second sample for each character lets the model route dialogue to the right mouth.

What to Avoid

✗

Collages and group grids

One subject per reference image. Crowded frames blur identity and wardrobe.

✗

Low-light or blurred stills

Soft input produces soft faces and drifting details in later shots.

✗

Overloaded prompts

Stacking ten actions into one 15-second clip forces the model to drop beats.

✗

Illegible brand assets

Small or low-resolution logos will not survive the render. Supply clean, high-resolution artwork.

Gemini Omni 1.2 Flash: Frequently Asked Questions

Model facts, availability, editing workflow and the limits Google still flags.

It is Google's omni-modal video model.

It takes text, images and video as input and returns video with synchronized audio, in clips of three to fifteen seconds.

Ready to Make Something Today?

Gemini Omni 1.2 Flash is coming to OnTrend AI. Until then, pick a trend-tested template, upload one or two photos, and post a finished clip in minutes.

Explore Viral Templates

No subscription · credits never expire

Gemini Omni is a trademark of Google LLC. OnTrend AI is an independent product and is not affiliated with, endorsed by, or sponsored by Google. Demo clips were published by Google for illustration and were not generated on OnTrend AI. Model capabilities and availability reflect Google's documentation as of November 2026 and may change.