Gemini Omni Flash 1.2 extends video generation to 15 seconds

• Video generation now reaches 15 seconds, with stronger camera controls and less visual drift, according to the launch announcement.
• Conversational edits preserve a performance while changing its setting, as demonstrated in a sequence of background swaps.
• Google cut 720p generation pricing 30% to $0.07 per second, as ozansihay reported. A 15-second output costs $1.05 at that rate.
• The model is available through partner platforms, including Runway and Magnific, alongside Google's developer tools.

A Google dialogue demo uses one-second voice references to steer an animated cast. Magnific's launch page accepts up to eight image references and three video references for generation.

Google began rolling out Gemini Omni 1.2 Flash on November 5. Its launch expands five parts of the video workflow:

• Duration: The maximum rises from 10 to 15 seconds. Length is selectable in one-second increments from 3 to 15 seconds.
• Camera control: Continuous flythroughs, zoom-outs and orbits get steadier scale and fewer unwanted jump cuts, Google says.
• Conversational editing: Follow-up prompts can revise a full 15-second clip, including replacing a product or changing the weather. Performance-preserving scene swaps are the standout upgrade.
• Consistency: Google reports better reference adherence and less morphing across frames, camera transitions and extensions.
• Price: The quoted 720p generation rate falls 30% to $0.07 per second. That puts a 15-second output at $1.05 before any additional editing or upscaling.

Google's launch examples, collected by minchoi, cover nine workflows:

1. Motion transfer: An octopus's rippling movement becomes the motion of a dancer's gown on the second editing turn.
2. Multi-speaker dialogue: A scene image, character and mouth references, and voice samples condition an animated conversation with distinct speakers and lip-sync.
3. Virtual try-on: One model photo plus front-and-back product shots of three outfits produce a 15-second wardrobe sequence, including turns and match cuts.
4. Background replacement: The setting changes through tent → forest → stable → market → castle across successive edits, retaining the characters' performances and camera framing.
5. Animated typography: A static exhibition poster becomes a five-layer animation, preserving the lettering HALO, GLOW and LUMEN through the motion.
6. Start-to-end frame interpolation: An exterior start frame and interior end frame guide a continuous camera push into a ramen bar.
7. Style-frame expansion: One style frame becomes a four-shot, 15-second science explainer with correctly spelled, motion-tracked callouts.
8. Architectural flythrough: Two reference photos guide a flight through the Pantheon, crossing the piazza, entering the bronze doors and rising toward the dome's oculus.
9. Extreme zoom-out: A shoreline close-up expands into an aerial island vista in one continuous 10-second shot.

Access spans several interfaces, with different controls and rollout conditions:

• Developer tools: Google AI Studio and the Gemini API are available, according to chrisfirst's launch post. Google's model documentation lists the model ID as gemini-omni-1.2-flash.
• Runway: The integration supports generating and editing clips up to 15 seconds, followed by 4K upscaling. A short walkthrough demonstrates text-prompt generation inside the platform.
• Magnific: Scripts and image/video references produce scenes with dialogue and sound generated together. Its interface offers Modify and Extend for subsequent revisions.
• Google apps: Google's launch post lists directing and multi-turn edits in Flow, plus a Gemini app rollout for Google AI Ultra subscribers.
• Google Vids: Anyone with a Google account or Workspace subscription gets up to three 15-second generations per month at no cost.

Availability varied by surface during the rollout. Ozansihay reported using the Gemini app, while an earlier reply said the model had not yet appeared for him in Flow.

Magnific's launch-day specifications list narrower editing controls than Google's full-length conversational editing feature:

• Modify duration: Input clips are limited to 3–10 seconds.
• Aspect ratio: Modify retains the original ratio, although Google's model-level announcement describes conversational aspect-ratio changes.
• Language: English is officially supported; other languages are not fully tested.
• Billing: Magnific uses its own credit rates, including 140 credits per second at 720p, rather than Google's quoted dollar rate.

Google also names three areas still under development in its launch post:

• Fast sports action.
• Rapid script changes.
• Complex physics.