Gemini Omni
Back to all articles
8 min read

Gemini Omni 1.2 Flash: 15-Second Clips, Multi-Turn Editing, and GA Pricing

Google made gemini-omni-1.2-flash generally available on November 5, 2026: 15-second clips, multi-turn editing, and about $0.07 per second of 720p video on the paid Gemini API tier.

Gemini OmniGemini Omni FlashOmni 1.2APIPricing2026

General availability on November 5

Google made Gemini Omni 1.2 Flash generally available on November 5, 2026. The model ID is gemini-omni-1.2-flash. Google describes this as the first time Gemini Omni is generally available.

Per the launch post, outputs carry an imperceptible SynthID watermark, and Google says Provisioned Throughput is available. Google’s media generation reference also lists C2PA metadata. The developer API has no free tier for Omni Flash.

If you still target gemini-omni-1.1-flash, plan for December 3, 2026, when the API routes 1.1 requests to 1.2 and bills at 1.2 rates (see AI Studio pricing and the model page).

15-second clips

Native generation now runs up to 15 seconds, up from 10 seconds on Omni 1.1. You can set duration in one-second steps from 3 to 15 seconds on supported modes.

Google Vids exposes a 3 to 15 second length slider. Per Workspace Updates, rollout to eligible Workspace and consumer accounts started November 5 and takes 1 to 3 days to appear. Google’s Vids post says anyone with a Google account or Google Workspace subscription can generate up to three 15-second videos per month at no cost in Vids.

Reference adherence, consistency, and speech

The launch post lists these changes:

AreaWhat Google highlights
ReferencesCharacter, product, or style images; short motion clips up to 3 seconds; audio cues
ConsistencyLess drift and morphing from first to last frame; steadier props and environments across transitions
SpeechMore natural dialogue, clearer lip-sync, distinct multi-speaker conditioning

On the API, you can attach up to 6 reference images for new generation, editing, and extension (including 15-second clips). Reference videos apply to new generation only: up to 3 references for 3 to 10 second clips and 1 for 11 to 15 second clips.

Uploaded source video limits depend on the task. The model page lists video up to 15 seconds for editing and 10 seconds lookback for extension. Treat the cap as up to 10 to 15 seconds depending on the task and confirm on the model page before you ship.

Multi-turn editing

Multi-turn conversational editing works across full 15-second clips. You pass previous_interaction_id on the Interactions API to swap products or backgrounds, change lighting or weather, or switch aspect ratio without regenerating from scratch.

Google’s workflow guidance: draft at 360p or 720p, fix details with a follow-up edit, then upscale the best take to 1080p or 4K.

Pricing

Published rates on the AI Studio pricing page (paid tier only):

MetricOmni 1.2 FlashOmni 1.1 Flash
Input (text, image, video) per 1M tokens$1.50$1.50
Output text (incl. thinking) per 1M tokens$9.00$9.00
Output video per 1M tokens$12.00$17.50
Effective 720p per secondabout $0.07about $0.10

Google says effective video output pricing dropped about 30% versus Omni 1.1. Billing uses 5,792 output tokens per second of 720p video at the $12 per 1M token video rate.

ResolutionApprox. $/second
360p$0.023
720p$0.07
1080p (with upsampling)$0.104
4K (with upsampling)$0.209
Clip (720p)Approx. cost
10 secondsabout $0.70
15 secondsabout $1.04

API limits by mode

Limits below follow the model page and Google’s Omni Flash API skill. All generation uses the Interactions API (client.interactions.create), not generate_content. Output is video with native audio plus text. 24 FPS. Aspect ratios: 16:9 and 9:16. Input cap 131,072 tokens; output cap 65,536 tokens.

ModeDurationReference notes
Text to video3 to 15 sImages up to 6; video refs per clip length rules above
Single first-frame image to video3 to 15 sSame
Reference to video and editing3 to 15 sImages up to 6
First and last frame interpolation3 to 10 sRequires both frames
Video extension per turn3 to 10 sUp to 30 s cumulative total

Resolutions: 360p, 720p (default), 1080p, 4K.

from google import genai

client = genai.Client()

interaction = client.interactions.create(
    model="gemini-omni-1.2-flash",
    input=[{"type": "text", "text": "A product demo on a clean stage."}],
    response_format={
        "type": "video",
        "duration": "10s",
        "resolution": "720p",
        "aspect_ratio": "16:9",
    },
)

Where it is available

SurfaceAccess
Gemini API and Google AI StudioPaid tier; starter apps and developer docs
Google Cloud console and Gemini for businessEnterprise solutions built on the API
Google FlowScenes and multi-turn edits; video extension requires Omni 1.2 and Google AI Plus, Pro, or Ultra (Flow help)
Google VidsUp to three 15-second videos per month at no cost with a Google account or Workspace subscription (Vids post)
Gemini app15-second videos rolling out to Google AI Ultra subscribers
Google AdsScale video campaigns

Deadlines for Omni 1.1 and Veo users

DateSurfaceChange
2026-11-05Google FlowVeo Extend retired; use Omni Flash 1.2 with AI Plus, Pro, or Ultra to extend (Flow help)
2026-12-03Gemini APIRequests to gemini-omni-1.1-flash route to gemini-omni-1.2-flash and bill at 1.2 rates
2026-12-05Google FlowVeo 3.1 Fast, Veo 3.1 Quality, and Omni Flash 1.1 retired; Veo 3.1 Lite stays (720p, 4/6/8 s, no video-to-video, no extend)

What Google says is still hard

Google notes that development continues on fast sports action, rapid script changes, and complex physics. Treat these cases as higher risk until Google ships further improvements.

Action checklist

  1. Point new API work at gemini-omni-1.2-flash on the Interactions API with paid-tier billing.
  2. Before December 3, 2026, regression-test any dependency on gemini-omni-1.1-flash pricing or clip limits.
  3. In Flow, migrate off Veo Extend and plan for December 5 retirements of Veo 3.1 Fast/Quality and Omni 1.1.
  4. Budget at about $0.07 per second of 720p output, using 360p drafts and edit passes before 1080p or 4K upscales.
  5. Read the launch post and model page before changing production prompts or reference bundles.