Oleksandr Kuzmenko, EditorPublished 5 min readAI-assisted · editor-reviewed
TL;DR
- Gemini Omni 1.2 Flash generates up to 15-second clips natively at $0.07 per second (720p).
- Multi-turn editing enables iterative adjustments to backgrounds or lighting without full re-generation.
- Model outputs are watermarked with SynthID and available via Google Cloud and Gemini API.
Key facts
- API Pricing
- $0.07 per second for 720p
- Cost Difference
- 30% reduction compared to Omni 1.1
- Maximum Duration
- 15 seconds (configurable from 3 to 15s)
- Watermarking
- SynthID imperceptible watermark
- Release Status
- General Availability via Gemini API
Conversational Video Refinements
Gemini Omni 1.2 Flash extends maximum clip length to 15 seconds, up from 10 seconds in Omni 1.1, with duration configurable in one-second increments between 3 and 15 seconds. The central technical update is multi-turn conversational editing: rather than generating an entirely new video to adjust an asset, developers submit follow-up prompts to modify specific properties such as backgrounds, lighting, weather, or aspect ratios while retaining character fidelity, camera position, and physical motions.
Reference Adherence and Audio Sync
The model accepts conditioning from multiple inputs simultaneously, including character sheets, motion clips of up to three seconds, and sample audio. In benchmarks showcased by Google, the system tracks typographic branding, garment drape, and facial geometry across 360-degree turns and continuous first-person flythroughs without unexpected jump cuts. Lip-sync conditioning enables distinct multi-speaker dialogue within a single 15-second scene.
Production Availability and Pricing
Now generally available on the Gemini API and Google Cloud console, Omni 1.2 Flash costs $0.07 per second for 720p output—a 30% discount relative to Omni 1.1. Generated content automatically receives an imperceptible SynthID watermark, and enterprise users can configure Provisioned Throughput for consistent latency.
When to use
- For social ads, e-commerce catalog 360-spins, and product explainers created from static reference photography.
- When an interactive workflow requires editing specific visual layers without regenerating the entire video.
When NOT to use
- For long-form video content generation exceeding 15 seconds without stitch transitions.
- For scenes requiring highly chaotic fast-action sports or complex real-world physics.
What to do today
- Test iterative video generation prompts in Google AI Studio before integrating the Gemini API.
- Adopt a draft-at-720p workflow to minimize token costs before upscaling selected clips.
- Verify your ingestion pipelines can handle SynthID metadata on generated outputs.
Step names are taken from What to do today.
Step names are taken from What to do today. Test iterative video generation prompts in Google AI Studio before integrating the Gemini API.. Adopt a draft-at-720p workflow to minimize token costs before upscaling selected clips.. Verify your ingestion pipelines can handle SynthID metadata on generated outputs.
- Test iterative video generation prompts in Google AI Studio before integrating the Gemini API.
- Adopt a draft-at-720p workflow to minimize token costs before upscaling selected clips.
- Verify your ingestion pipelines can handle SynthID metadata on generated outputs.