Google’s Gemini Omni 1.2 Flash Makes 15-Second Videos for 30% Less. What About Arabic?

On 5 Nov 2026, Google DeepMind launched Gemini Omni 1.2 Flash (model id gemini-omni-1.2-flash), making native audio and video generation generally available across the Gemini API and Google Cloud. For creative teams monitoring the wider Gemini ecosystem, the release cuts the price by 30% and stretches clips to 15 seconds. Our verdict: the cheapest way to try it is Google Vids, at no additional cost, and that is where you should start. Anyone making videos for Arab audiences should hold off on paid API work, because Google has not evaluated Arabic at all.

Longer Clips and Conversational Edits

Where Omni 1.1 capped output at 10 seconds, version 1.2 generates clips up to 15 seconds, as The Decoder also reported. In the Gemini API docs, developers can configure clip durations from 3 to 15 seconds in 1-second steps, with a default duration of 10 seconds in either 16:9 or 9:16 aspect ratios. Clip extensions run 3 to 10 seconds per turn up to a maximum of 30 seconds total by appending strictly to the end, while first-to-last frame interpolation spans 3 to 10 seconds.

As detailed on the Google blog, the headline workflow improvement is multi-turn conversational editing across full 15-second generations. Creators can swap out backgrounds, replace featured products, adjust lighting conditions, or switch aspect ratios using follow-up text prompts. The API supports reference inputs including up to 6 reference images (up to 3 per character for up to 2 characters) and short motion clips up to 3 seconds. However, audio reference uploads remain unsupported in the public API and are restricted to an allowlist on Google Cloud.

According to Google, the new model provides better consistency for faces, clothing, brand colours, and packaging typography, alongside multi-speaker lip-sync and smoother camera pans. Nobody has tested those claims independently yet. Google also openly acknowledges it is still working to resolve fast sports action, rapid script shifts, and complex physics.

API Pricing and Resolution Tiers

Per-second pricing is what decides whether AI video works for a business, and it is where Google is now leaning on rivals like Runway and Luma. On the Gemini API pricing page, standard 720p generation costs $0.07 per second—30% below Omni 1.1’s rate of ~$0.10 per second—billed as output tokens at $12 per 1M tokens. From 3 Dec 2026, API requests to gemini-omni-1.1-flash route automatically to 1.2 Flash at 1.2 rates. Note that the API offers this model strictly under a paid tier; a free tier is not available.

To keep production costs low, Google recommends drafting at 360p or 720p, fixing flaws with conversational edits, and upscaling only the final selected clip to 1080p or 4K.

Platform Availability and Regional Rules

Everyday users can test 15-second generation at no additional cost in Google Vids using a personal Google account or Workspace subscription. DeepMind also rolled the model into Flow for multi-turn scenes, Google Ads for automated campaigns, and Google AI Studio for developers. In the Gemini app, access to 15-second generation is rolling out exclusively to Google AI Ultra subscribers, while the Google Cloud blog notes it is generally available on Google Cloud for the first time.

Every rendered clip embeds an invisible SynthID watermark alongside C2PA Content Credentials, according to Google. The Gemini API docs detail regional safety restrictions: uploading images of certain recognizable individuals is barred. In the UK, EEA, and Switzerland, editing or extending uploaded video and editing images of minors are disabled, and safety filters vary regionally.

The Arabic Blind Spot

For agencies in the region already using tools like Google Veo, Google’s language note is the real obstacle. The Gemini API docs state that English is fully supported, but other languages “have not been evaluated” and may yield inconsistent results. Across Google’s entire announcement, documentation, and pricing sheets, the word Arabic never appears once. The sole non-English language acknowledged across launch materials is Punjabi dialogue in a partner testimonial on DeepMind’s model page. Furthermore, Google has not published regional pricing in SAR, AED, or EGP, and has said nothing specific about the Middle East.

What to Do Now: The 10-Minute Diagnostic

Do not rebuild commercial production pipelines based on vendor demo reels. Because Arabic is unevaluated, run a quick diagnostic test inside Google Vids at zero cost before committing API spend.

Generate a 15-second test video depicting a commercial storefront with visible Arabic signage, featuring a character speaking a single line in Egyptian or Gulf dialect. Inspect the output across four specific checkpoints:

• Arabic Typography: Confirm whether letters join correctly from right to left or render disjointed and reversed on signs and packaging.
• Phonetic Lip-Sync: Verify that mouth movements follow distinct Arabic consonants and vowels rather than a generic rhythm that doesn’t match the words.
• Dialect Naturalness: Listen closely for unnatural robotic cadences or awkward pronunciation in colloquial phrasing.
• Scene Stability: Verify that facial features, lighting, and brand colours stay consistent across camera movement without warping.

If errors appear, use conversational prompts to see whether follow-up edits can correct them without full regeneration. Until Google officially evaluates Arabic, treat Gemini Omni 1.2 Flash as an English-first prototyping tool with attractive economics, but unverified regional fundamentals.