Gemini Omni 1.2 Flash: Videos Up to 15 Seconds and API Prices Drop

Key Takeaways

1Gemini Omni 1.2 Flash allows for the generation of videos ranging from 3 to 15 seconds

2Conversational editing enables text-guided, step-by-step adjustments

3The 720p API rate decreases by 30%, set at $0.07 per second

4Google integrates SynthID and acknowledges limitations on certain movements

Why it matters : This model aims to make video creation more accessible and flexible, while showcasing advancements in visual consistency and cost management.

Google launches Gemini Omni 1.2 Flash, a video model focused on conversational editing and the generation of clips up to 15 seconds long. The API price for 720p decreases by 30% to $0.07 per second, with integration into the Google ecosystem and announced technical limits and safeguards.

Recognized Limitations and Safeguards Activated by Google

According to Google, Gemini Omni 1.2 Flash still has limitations when it comes to accurately rendering fast sports movements, sudden changes, and certain complex physical interactions. Therefore, a video that appears credible at first glance may contain inconsistencies. To facilitate the identification of AI-generated content, Google integrates the invisible digital watermark SynthID. The model's technical specifications also indicate that altering people's speech remains subject to restrictions. Additionally, the visual continuity worked on with Genie 3 is mentioned, although the objectives differ from those of Gemini Omni 1.2 Flash.

Declining 720p Costs and Suggested Working Method

Google announces a 30% reduction in the video generation rate for 720p via its API, now set at $0.07 per second. At this price, creating a 15-second clip costs $1.05. Google advises designing the clip in 360p or 720p, making necessary corrections, and then increasing the resolution of the final render to 1080p or 4K.

Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisissez votre rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

Conversational Editing and Successive Touch-ups

Gemini Omni 1.2 Flash allows for conversational touch-ups in natural language. Users can request to change the background, alter the weather, adjust the lighting, or adapt the video format. Conversational editing occurs in several steps, with each instruction refining the previous result. This approach already existed in the Gemini Omni ecosystem, particularly in Google Vids, but the novelty lies in the precision of the touch-ups and their support for longer sequences.

Generation: 3 to 15 Seconds, Enhanced Consistency, and Improved Voices

Gemini Omni 1.2 Flash can generate videos ranging from 3 to 15 seconds, whereas the previous version was limited to 10 seconds. Google claims that the model offers better visual consistency on faces, clothing, and objects when the camera moves, relying on reference images or short video clips. Dialogues benefit from more natural voices and improved lip synchronization, continuing the advancements already introduced in Google Vids. On the Arena platform, Gemini Omni 1.2 Flash ranks first for video generation from text and images.

Integrations, Professional Use Cases, and Early Users

Gemini Omni 1.2 Flash is available via the Gemini API and Google Cloud, and integrates with Google Vids. The Gemini application plans a gradual rollout for AI Ultra subscribers. Google cites professional applications such as personalized advertisements, product presentations, and virtual try-ons. Adobe, Monks, and Whering are mentioned among the early users. The model allows for the generation of clips up to 15 seconds long.