Gemini Omni 1.2 Flash makes a 15-second AI clip for $1.04, if you buy it from Google

Google made Gemini Omni 1.2 Flash generally available on Thursday on the Gemini API (Google's paid service for developers) and Gemini for business. Clips now run up to 15 seconds, up from 10, you can edit a clip over several chat turns, and the API price drops 31%, to about $0.07 a second at 720p, or $1.04 for 15 seconds.

That makes Omni the cheapest major video model at 720p on list price this week, 26% under Kling 4.0 Flash, which opened to everyone a day earlier. But the number in the headlines is Google's paid developer rate. Runway had the model live within the hour, and its help center still documents a rate roughly three times higher. Editing a clip instead of re-rolling it isn't free either, because every edit turn renders and bills a whole new clip. Where you buy Omni matters more than the 30%.

What Google shipped

Google's launch post, written by Google DeepMind product manager Anish Nangia, and the Gemini API changelog both date the release to Thursday 5 November. Clips run from 3 to 15 seconds in one-second steps at 24 frames a second, and the extension feature can carry a shot to 30 seconds. Some older limits stay. Interpolating between a first and a last frame still stops at 10 seconds, and clips longer than 10 seconds take one reference video of up to 3 seconds, against three for shorter clips. Up to six reference images work at any length.

Google calls this the first time Omni is generally available. That holds for Google Cloud, but the Gemini API already made Omni 1.1 generally available on Thursday 27 August. Two details most of the coverage skipped matter more to anyone building on it. From Thursday 3 December, every API call to Omni 1.1 routes to 1.2 and bills at 1.2 rates, with no code change. And since Google shut down Veo 3.1 Preview in the Gemini API on 22 October, Omni is now the only video model in Google's developer API.

The price sheet, line by line

Google bills video as output tokens, at $12 per million, down from $17.50 on Omni 1.1. A second of 720p video counts as 5,792 tokens, which is where the $0.07 comes from. Read the other rows closely: 1080p and 4K are upscaled from the native render, not generated at that size, and they cost one and a half and three times the 720p rate.

• Output (Gemini API list, read Fri 6 Nov 2026): 360p draft. Omni 1.2 per second: $0.023; Omni 1.1 per second: $0.034; 15-second clip on 1.2: $0.35
• Output (Gemini API list, read Fri 6 Nov 2026): 720p, the default. Omni 1.2 per second: $0.0695; Omni 1.1 per second: $0.101; 15-second clip on 1.2: $1.04
• Output (Gemini API list, read Fri 6 Nov 2026): 1080p, upscaled. Omni 1.2 per second: $0.104; Omni 1.1 per second: $0.152; 15-second clip on 1.2: $1.56
• Output (Gemini API list, read Fri 6 Nov 2026): 4K, upscaled. Omni 1.2 per second: $0.209; Omni 1.1 per second: $0.304; 15-second clip on 1.2: $3.13

There is no free API tier, and the rate limits page lists free use as not supported. The free route is Google Vids, where Google says anyone with a Google account can make three 15-second videos a month at no cost. In the Gemini app, Google says 15-second clips are rolling out to its AI Ultra plan first, and multi-turn edits to Google AI subscribers.

The day-one apps aren't at Google's price

Runway announced the model at 17:54 UTC on Thursday, 54 minutes after Google's post, promising videos "up to 15 seconds" and an upscale to 4K. Krea followed about eight hours later. Neither published a price for 1.2. Runway's own help article for Gemini Omni Flash, read on Friday morning, still describes the 1.1 setup: 3 to 10 seconds, 720p only, a Standard plan or higher, and 10 credits a second. On Runway's Standard plan, which gives 625 credits for $12 to $15 a month, that works out to about $0.19 to $0.24 a second, roughly three times Google's list.

We don't know yet whether 1.2 keeps that rate in Runway, so treat it as the documented price rather than a confirmed one. Krea doesn't publish a per-model cost at all. Developer resellers sit close to list, and OpenRouter says its price is the same as going direct to Google.

Every edit is a new render, and it's billed like one

Multi-turn editing is the real workflow change. You keep a clip you like and ask for one change, a different jacket or a slower camera move, without rolling the dice on the whole shot again. Google bills on total output tokens, though, and each edit returns a full clip. Runway's developer docs for Omni 1.1 say an edit "is billed on the length of its source clip". Our reading is that changing one detail in a 15-second 720p clip costs about another $1.04. The saving is in the good takes you stop throwing away, not in cheaper turns.

The feature also arrived unevenly. Segmind, a reseller that ran 20 test jobs on launch day, found its multi-turn endpoint answering that it was not available yet. Some of Google's own demos lean on inputs the API doesn't take: the four-character dialogue scene uses short voice references, and the API guide says audio inputs are "unsupported in the current version of the API". In the UK, the EEA and Switzerland, editing or extending footage you upload is blocked, though editing Omni's own clips works.

Cheap isn't the same as best

There is no independent quality score for 1.2 yet. Artificial Analysis has no row for it, and Omni 1.1 sits 10th on its text-to-video board. Posts on X say 1.2 went straight to first place on LMArena, but we couldn't open the board to check. Creators posting side-by-side tests mostly land on the same split: Seedance 2.5 still wins on photoreal people and motion, while Omni wins on price and on doing what the prompt asks. Segmind's review called it "the most instruction-exact video model I have tested" and also "not a cinematic, photoreal-people model". Google says it is still working on fast sports action, rapid script changes and complex physics.

Our take

Omni 1.2 Flash is now the cheapest serious engine for drafting a scene. Block a shot at 360p for 35 cents, then pay $1.04 for the 720p take that works, and a short gets affordable fast. It suits explainers, product shots, type and stylized looks, and it isn't yet the model for photoreal acting. Buy it where the price is published, and price each edit turn as a full render before you build a workflow around talking to your footage.

Timeline

• Date (UTC): Tue 30 Jun 2026. Event: Omni Flash preview released, 3 to 10 seconds at 720p; Source: Gemini API changelog
• Date (UTC): Thu 27 Aug 2026. Event: Omni 1.1 Flash generally available in the Gemini API at $17.50 per 1M tokens; Source: Gemini API changelog
• Date (UTC): Thu 22 Oct 2026. Event: Veo 3.1 Preview shut down in the Gemini API; Source: Gemini API video page
• Date (UTC): Wed 4 Nov 2026. Event: Kling 4.0 Flash opens to everyone at $0.094 a second, 720p; Source: Kling on X; Kling price list
• Date (UTC): Thu 5 Nov 2026, 17:00. Event: Omni 1.2 Flash generally available, 15 seconds, $12 per 1M tokens; Source: Google blog; Gemini API changelog
• Date (UTC): Thu 5 Nov 2026, 17:54. Event: Live in Runway; Source: Runway on X
• Date (UTC): Thu 5 Nov 2026, 20:00. Event: Google Vids moves to 1.2 with three free 15-second videos a month; Source: Google Workspace blog
• Date (UTC): Fri 6 Nov 2026, 01:00. Event: Live in Krea; Source: Krea on X
• Date (UTC): Thu 3 Dec 2026. Event: All Omni 1.1 API calls route to 1.2 at 1.2 rates; Source: Gemini API pricing page

Sources and what is unconfirmed: Google's launch post and the Google DeepMind post of 5 November; the Gemini API changelog, pricing page, rate limits page, Omni model page and Omni guide, read on 6 November; the Google Workspace post on Vids; Runway's and Krea's posts on X; Runway's help article, pricing page and developer docs, read on 6 November; OpenRouter on X; Segmind's launch-day review; the Artificial Analysis text-to-video board, read on 6 November; Kling's price list, read on 5 November; X API searches on 5 and 6 November. Unconfirmed: Runway's and Krea's prices for 1.2, the LMArena ranking, render times, and whether edits hold a character across turns.

Your Voice

Quoted from public posts on X and credited by handle. If one of these is yours and you want it removed, message @the__generated on X and it comes down.

Got something to say? Tag @the__generated on X