Give the scene a starting point.
Use visual references to guide the scene. Google reports improved reference adherence.
✦ Google DeepMind · Model guide
Gemini Omni 1.2 Flash
More room for your next scene.
Longer video. More natural dialogue. A closer look at Google’s latest Omni model, with its official demonstrations.
Model introduction only. No Offloop integration is announced.
A little world. A conversation.
Google’s published character-and-voice demonstration. The current Gemini API does not accept uploaded audio references.
- Native generation, up to
- 15 seconds
- Google-reported vs Omni 1.1
- 30% lower price
- Google API · 720p video output
- ≈$0.07 / second · USD
Google’s November 5 announcement reports a 30% reduction. Standard API video output is $12 per million tokens; 720p uses 5,792 tokens/second (about $0.0695/second). Input and text-output charges are separate. This is not Offloop pricing. Google pricing ↗
Direct the idea. Then refine it.
A useful creative process leaves room for another look. Start with references, make a first version and decide what the next edit should change.
Keep the conversation going.
Revise clips through follow-up instructions.
Think in shots, not just frames.
Improved camera control and speech support richer scenes.
Capabilities described by Google. The examples above were published by Google, not produced in Offloop.
A shared place for your next brief.
Offloop brings people, AI Agents and project context into the same workspace. Use it to organize the discussion around your next creative project.
01 / BRIEF
Agree on the direction.
Keep the audience, references and acceptance criteria together.
02 / REVIEW
Make feedback specific.
Discuss what to keep, what to change and who makes the call.
03 / HANDOFF
Leave the context intact.
Give the next person the decisions behind the finished brief.
A few things to know.
Can I use this model in Offloop?
This guide does not confirm model availability in Offloop. For access to Gemini Omni 1.2 Flash, consult Google’s official services; the Gemini API lists it on the paid tier.
Are these Offloop-generated videos?
No. All four demonstrations are Google’s original published media, streamed from its official asset host. They show selected examples, not guaranteed results.
Does every video last 15 seconds?
Ordinary generation and editing support 3–15 seconds, with a 10-second default. First/last-frame interpolation and each extension step support 3–10 seconds; extended videos can total up to 30 seconds. Published demonstrations may include multiple segments. Duration rules ↗
Can I upload a voice reference?
Google’s release demonstration uses voice conditioning. Its current Gemini API documentation says uploaded audio inputs and audio references are unsupported; voice editing is also unsupported. Native audio generation is a separate capability. API limitations ↗
Where can I verify the information?
Read Google’s November 5 announcement, the Google AI Studio post, and its developer documentation. Official examples may still show generation imperfections.
Your next project starts with a good brief.
Learn about OffloopThis guide does not announce Gemini Omni 1.2 Flash availability in Offloop.