Gemini Omni 1.2 Flash vs Gemini Omni 1.1 Flash: Complete Comparison and Report on Price per Second, Clip Length, Upscaling Costs, and Reference Control
Gemini Omni 1.2 Flash is the second generally available version of Google's Omni video generation and editing model, announced by Google DeepMind on November 5, 2026 and declared generally available on Google Cloud the following day. It replaces Gemini Omni 1.1 Flash as the model Google promotes across the Gemini API, Google Cloud, Flow, Google Vids and the Gemini app, while the older model, released on August 27, 2026 under the API identifier gemini-omni-1.1-flash, still appears as the stable model on the Gemini API model page and still carries its own price on Google Cloud's pricing page.
The verdict on the facts Google has published is clear for anyone paying per second. Omni 1.2 Flash costs $12.00 per million video output tokens against $17.50 for Omni 1.1 Flash, a 31.4% lower rate that applies identically at 360p, 720p, 1080p and 4K, so a second of 720p video falls from about $0.101 to about $0.070. The newer model also generates clips of up to 15 seconds in one pass, against a 10-second ceiling on 1.1, adds audio references (behind an allowlist) and multi-speaker conditioning, and is the version now serving video in the Gemini app. The one measurable advantage left to Omni 1.1 Flash is an independent quality ranking, which 1.2 does not yet have. Artificial Analysis lists Omni 1.1 Flash at an Elo of 1127 in its text-to-video arena, in twelfth place, while Omni 1.2 Flash has no public score from any evaluator and Google has published no quantitative benchmark for it.
For developers and businesses generating video at volume, Omni 1.2 Flash is the better choice on price and clip length. Omni 1.1 Flash remains defensible only for teams with a validated production pipeline that depends on its documented behavior and that want to wait for independent quality data on 1.2 before switching. Prices in this comparison were verified on Google's official pricing page on November 9, 2026.
··········
GEMINI OMNI 1.1 FLASH AND 1.2 FLASH: RELEASE TIMELINE AND MODEL IDENTITY.
Google introduced Gemini Omni on May 19, 2026 as a model that creates video from any combination of text, images, audio and video, with the first release, Gemini Omni Flash, rolling out to Google AI Plus, Pro and Ultra subscribers in the Gemini app and Flow, and at no cost in YouTube Shorts and the YouTube Create app. Developer access followed on June 30, 2026, when the Gemini API release notes list gemini-omni-flash-preview as a public preview producing 3 to 10 second clips at 720p with conversational editing through the Interactions API.
Omni 1.1 Flash was the production release. The Gemini API release notes of August 27, 2026 describe gemini-omni-1.1-flash as the generally available version of the model and list three new capabilities: video extension, first-and-last-frame interpolation with up to two images, and a resolution parameter offering 360p, 720p (the default), 1080p and 4K, with the two higher resolutions produced by upscaling. The same entry scheduled the preview endpoint for deprecation on September 30, 2026. Google Vids received Omni 1.1 Flash on September 23, 2026, with 1080p scenes available to anyone with a Google or Workspace account.
Omni 1.2 Flash arrived ten weeks after 1.1. Google DeepMind's announcement is dated November 5, 2026 and Google Cloud's general availability post November 6, 2026.
........
Milestone | Gemini Omni 1.1 Flash | Gemini Omni 1.2 Flash |
Announcement | August 27, 2026 | November 5, 2026 |
Google Cloud status | Listed with its own price | Generally available, November 6, 2026 |
API model ID (official) | gemini-omni-1.1-flash | Not shown on a page fetched for this report |
Predecessor | gemini-omni-flash-preview (June 30, 2026) | Gemini Omni 1.1 Flash |
Retirement date | None published | Not applicable |
Gemini app status | Replaced by 1.2 in the app help page | Current app model |
........
The Gemini API model page, last updated October 29, 2026, still lists gemini-omni-1.1-flash as the stable model with no deprecation date, and the Gemini API release notes read for this report end on October 26, 2026 with no entry for 1.2. Google's announcement and its Cloud pricing page are therefore the primary sources for the new version.
··········
GEMINI OMNI 1.2 FLASH VS 1.1 FLASH PRICING PER TOKEN AND PER SECOND.
Both models are billed in tokens on Google Cloud, and the only rate that differs is video output. Input of any kind (text, images, video, audio) costs $1.50 per million tokens on both, and text output, which covers responses and reasoning, costs $9.00 per million tokens on both. Video output costs $12.00 per million tokens on Omni 1.2 Flash and $17.50 on Omni 1.1 Flash.
The pricing page converts video into tokens at fixed rates: 1,931 tokens per second at 360p, 5,792 at 720p, 8,688 at 1080p and 17,376 at 4K, all including audio. Multiplying those rates by the per-token prices gives the per-second costs below, computed in Python. Google's own headline figure for Omni 1.2 Flash, $0.07 per second at 720p, is the rounded value of $0.0695.
........
Output resolution | Tokens per second | Omni 1.1 Flash per second | Omni 1.2 Flash per second | Omni 1.1 Flash per minute | Omni 1.2 Flash per minute |
360p | 1,931 | $0.0338 | $0.0232 | $2.03 | $1.39 |
720p | 5,792 | $0.1014 | $0.0695 | $6.08 | $4.17 |
1080p | 8,688 | $0.1520 | $0.1043 | $9.12 | $6.26 |
4K | 17,376 | $0.3041 | $0.2085 | $18.24 | $12.51 |
........
Because the token rates per second are the same for both versions, the saving is a constant 31.4% at every resolution. Google rounds it to 30% and the Google Cloud post says the cut applies at 720p "with similar savings at other resolutions", which the token arithmetic confirms exactly. The $9.12 per minute that Artificial Analysis shows for Omni 1.1 Flash on its leaderboard matches the 1080p rate computed from Google's page, which suggests the evaluator prices the model at 1080p.
Input costs are small next to output. An image reference is counted as 1,120 tokens, or $0.00168, and a second of audio input as 32 tokens, or $0.000048. A second of uploaded video for editing or extension counts as up to 5,792 tokens, or up to $0.0087, on either model.
··········
UPSCALING COSTS AND THE 360P DRAFT PATH ON BOTH OMNI VERSIONS.
Omni produces 1080p and 4K by upscaling, and Google prices the upscaling step separately: 2,896 output tokens per second from 360p to 720p and from 720p to 1080p, 5,792 tokens per second from 360p to 1080p, and 11,584 tokens per second from 720p or 1080p to 4K. Upscaling input is counted at zero tokens.
The rates line up with the direct prices. A 720p generation followed by a 720p-to-1080p upscale adds up to 8,688 tokens per second, the same as a direct 1080p output, and a 720p generation followed by a 4K upscale adds up to 17,376, the same as direct 4K. A 360p generation followed by an upscale is cheaper per second than the equivalent direct output, as the table shows.
........
Path to the final clip | Tokens per second | Omni 1.1 Flash per second | Omni 1.2 Flash per second |
720p direct | 5,792 | $0.1014 | $0.0695 |
360p then upscale to 720p | 4,827 | $0.0845 | $0.0579 |
1080p direct, or 720p then upscale | 8,688 | $0.1520 | $0.1043 |
360p then upscale to 1080p | 7,723 | $0.1352 | $0.0927 |
4K direct, or 720p then upscale | 17,376 | $0.3041 | $0.2085 |
Upscale only, 720p or 1080p to 4K | 11,584 | $0.2027 | $0.1390 |
........
Google designed both versions around this draft workflow. The Omni 1.1 Flash announcement describes 360p as a lightweight preview mode that is up to 60% faster than 720p, by Google's throughput measurement, at a third of the cost. The Omni 1.2 Flash announcement keeps the same structure, with drafts at 360p or 720p and finished clips upscaled to 1080p or 4K. The cheaper 360p-then-upscale path in the table is pure arithmetic from Google's token rates; Google has published no comparison of the visual quality of an upscaled 360p draft against a direct 720p generation on either model.
··········
CLIP LENGTH, RESOLUTION AND EDITING LIMITS: OMNI 1.2 FLASH VS 1.1 FLASH.
The single largest functional change is duration. Omni 1.1 Flash generates clips of 3 to 10 seconds at 24 frames per second, according to the Gemini API model page, and extends them in 10-second increments up to a cumulative 40 seconds. Omni 1.2 Flash generates and edits clips of up to 15 seconds, with duration set in one-second steps from 3 to 15 seconds.
........
Specification | Gemini Omni 1.1 Flash | Gemini Omni 1.2 Flash |
Single-generation length | 3 to 10 seconds | 3 to 15 seconds, in 1-second steps |
Extension | 10-second increments, up to 40 seconds total | Not specified in the announcement |
Native draft resolutions | 360p, 720p (default) | 360p, 720p |
Upscaled resolutions | 1080p, 4K | 1080p, 4K |
Frame rate | 24 FPS | Not stated |
Aspect ratios (Gemini API) | 16:9 (default), 9:16 | Not stated |
Video references | Up to 3 clips of up to 3 seconds each | Motion reference clips of up to 3 seconds each |
Audio references | Unsupported in the API | Supported, behind an allowlist |
Context window | 1,048,576 tokens | Not stated |
Watermark | SynthID | SynthID |
........
Omni 1.1 Flash's Gemini API guide documents the limits developers work with today: uploaded videos for editing or extension must be 10 seconds or shorter, extension appends only to the end of a clip, uploaded videos cannot be extended with spoken dialogue, voice editing is unsupported, and system instructions, temperature, top_p, stop sequences and negative prompts are not accepted. Outputs are MP4 files returned in base64, with URI delivery recommended above 4 MB. The guide states that English is fully supported and that other languages have not been evaluated.
For Omni 1.2 Flash, Google's announcement describes multi-turn editing across the full 15-second clip and says motion reference clips remain limited to three seconds each. The 1.2 entry in Google Cloud's documentation could not be read in full for this report, so the specification cells marked "not stated" reflect what the fetched official pages say, and they may be filled in Google's model documentation.
··········
REFERENCE ADHERENCE, SPEECH AND CONSISTENCY: WHAT GOOGLE CLAIMS FOR OMNI 1.2 FLASH.
Google lists five qualitative improvements of Omni 1.2 Flash over 1.1. The model follows image and video references more closely, keeping faces, garments, packaging typography and brand colors consistent; it holds characters, objects and scenes steadier, with less drift or morphing; it produces more natural speech with clearer lip-sync and distinct multi-speaker conditioning; it inserts fewer unwanted jump cuts on complex camera moves; and it supports multi-turn edits across the whole clip.
Each of these claims is illustrated in Google's announcement with side-by-side examples. None is accompanied by a measured result, an evaluation set, a human-preference rate or a comparison score. Google also lists where the model still needs work: fast sports action, rapid script changes and complex physics.
Customer statements in the Google Cloud post describe the same improvements from the users' side. Photoroom's chief executive calls 15-second videos "a top user request" and says general availability with service-level agreements lets the company rely on the model in production. QuickFrame AI says 15-second generations at 1080p allow a complete connected-TV advertisement in a single generation. Picsart, which cites more than 100 million creators, describes iterating in draft mode and rendering in 4K. These statements were selected by Google for its own announcement and are vendor-curated testimony, separate from independent evaluation.
Omni 1.1 Flash brought its own set of capabilities in August: scene extension that reads up to 10 seconds of prior context, where earlier models referenced only the final second; first-and-last-frame interpolation; camera-control prompting; and video references for character consistency. Those capabilities carry into 1.2 by Google's description of the model as an upgrade of the same family.
··········
[[ADPLUS_300x250_1]]
INDEPENDENT EVALUATIONS OF GEMINI OMNI FLASH: ARTIFICIAL ANALYSIS RESULTS.
The only independent quality measurement found for either model is the Artificial Analysis text-to-video arena (AA-Video-T2V v2.1, all prompts, with audio), where users vote between pairs of clips and the results are converted into Elo ratings anchored on Kling 3.0 1080p (Pro) at 1000. On the leaderboard read on November 9, 2026, Gemini Omni Flash 1.1 ranks twelfth with an Elo of 1127, a 95% confidence interval of 1114 to 1140, and 7,807 samples. Gemini Omni 1.2 Flash does not appear on the leaderboard.
........
Model (Artificial Analysis naming) | Creator | Rank | Elo | 95% interval | Price per minute (AA) |
Wan 3.0 | Alibaba | 1 | 1172 | 1159 to 1185 | $12.00 |
MiniMax H3 (768p) | MiniMax | 3 | 1160 | 1147 to 1173 | $4.80 |
Dreamina Seedance 2.5 | ByteDance Seed | 5 | 1155 | 1142 to 1168 | $34.12 |
Gemini Omni Flash 1.1 | 12 | 1127 | 1114 to 1140 | $9.12 | |
Grok Imagine Video 1.5 | SpaceXAI | 14 | 1055 | 1041 to 1069 | $15.00 |
Veo 3.1 | 23 | 963 | 948 to 978 | $24.00 | |
Gemini Omni 1.2 Flash | Not listed | Not rated | Not rated | Not listed |
........
Two readings follow from the table and stay within what it measures. Omni 1.1 Flash already ranks well above Google's own Veo 3.1 on this arena, by 164 Elo points, at a lower listed price per minute. It also sits about 45 points below the leader, Wan 3.0, and outside the top ten. Whether Omni 1.2 Flash closes that gap is an open question until Artificial Analysis or another evaluator rates it. The arena measures preference on text-to-video prompts; it does not measure reference adherence, editing or lip-sync, which are the areas Google says improved most.
··········
ACCESS TO GEMINI OMNI 1.2 FLASH AND 1.1 FLASH BY PLATFORM AND PLAN.
Omni 1.2 Flash takes over the surfaces where 1.1 was offered, and it adds Google Ads. Google DeepMind's announcement lists the Gemini API in Google AI Studio, the Google Cloud console and enterprise API with Provisioned Throughput, Flow, Google Vids, the Gemini app and Asset Studio in Google Ads. The Google Cloud post adds that audio references are gated behind allowlisting and describes Gemini for business access as coming soon, while the DeepMind post describes the model as generally available on Gemini for business; the two official posts disagree on that point.
........
Surface | Gemini Omni 1.1 Flash | Gemini Omni 1.2 Flash |
Gemini API (Google AI Studio) | Stable model, gemini-omni-1.1-flash | Generally available per Google DeepMind |
Google Cloud (Agent Platform) | Priced at $17.50 per 1M video tokens | Generally available, $12.00 per 1M video tokens |
Provisioned Throughput | Not supported in the Gemini API guide | Offered on Google Cloud |
Gemini app | Previous model | Current model; 15-second clips for Google AI Ultra |
Flow | Google AI Plus, Pro and Ultra subscribers | Available for scene direction and multi-turn edits |
Google Vids | Added September 23, 2026, 1080p, no cost | 15-second clips at no additional cost |
Google Ads Asset Studio | Not announced | Available for video campaigns |
Gemini for business | Not announced | Described as available and as coming soon |
........
In the Gemini app, the help page now names Omni 1.2 Flash as the latest video model and sets clip length at 3 to 10 seconds, or up to 15 seconds for Google AI Ultra subscribers. Each editing turn accepts one video and up to five images, a generated clip can be extended by up to another 10 seconds, and the feature is unavailable to users under 18. Uploading videos for editing and creating avatars are unavailable in the EEA, Switzerland, the United Kingdom and some US states. The Gemini Apps limits page marks video generation as available on Google AI Plus, Pro and Ultra, rates it as the most usage-intensive feature, and scales limits at 2x the standard level for Plus, 4x for Pro and 5x or 20x the Pro level for Ultra. Our comparison of Google AI Plus and Google AI Pro covers those plans in detail, and our guide to Google Vids explains its free and paid tiers.
··········
COST OF THREE TYPICAL VIDEO WORKLOADS ON OMNI 1.2 FLASH AND 1.1 FLASH.
The three workloads below use only the token rates and prices on Google Cloud's pricing page and count video output tokens alone, since input tokens for a text prompt are negligible and Google publishes no per-request fee. Workload one is a single 15-second 720p clip; on Omni 1.1 Flash it requires a 10-second generation plus an extension, and the figure counts only the 15 seconds of output, leaving out the input tokens of the extension context. Workload two is a 10-second advertisement made from eight 360p drafts, one 720p final and a 720p-to-4K upscale. Workload three is a catalog of 500 ten-second product videos at 1080p per month, computed both as direct 1080p output and as 360p drafts upscaled to 1080p.
........
Workload | Omni 1.1 Flash | Omni 1.2 Flash | Difference |
One 15-second 720p clip (output only) | $1.52 | $1.04 | $0.48 less on 1.2 |
10-second ad: 8 drafts at 360p, 720p final, 4K upscale | $5.74 | $3.94 | $1.80 less on 1.2 |
500 ten-second clips at 1080p, direct | $760.20 | $521.28 | $238.92 less on 1.2 |
500 ten-second clips, 360p upscaled to 1080p | $675.76 | $463.38 | $212.38 less on 1.2 |
........
The calculation in Python follows Google's published rates.
PRICE = {"1.2": 12.00, "1.1": 17.50} # USD per 1M video output tokens
TOKENS = {"360p": 1931, "720p": 5792, "1080p": 8688, "4K": 17376}
UP_720_TO_4K, UP_360_TO_1080 = 11584, 5792 # upscaling tokens per second
def cost(model, tokens_per_s, seconds):
return tokens_per_s * seconds * PRICE[model] / 1e6
for m in ("1.1", "1.2"):
clip15 = cost(m, TOKENS["720p"], 15)
ad = cost(m, TOKENS["360p"], 8 * 10) + cost(m, TOKENS["720p"], 10) + cost(m, UP_720_TO_4K, 10)
catalog = cost(m, TOKENS["1080p"], 500 * 10)
catalog_draft = cost(m, TOKENS["360p"] + UP_360_TO_1080, 500 * 10)
print(m, round(clip15, 2), round(ad, 2), round(catalog, 2), round(catalog_draft, 2))The draft-heavy advertisement shows where the saving accumulates: eight 360p drafts cost $2.70 on Omni 1.1 Flash and $1.85 on Omni 1.2 Flash, so a team that iterates many times before rendering a final saves on every discarded draft. The catalog case, at 5,000 seconds of finished video a month, saves $238.92 on direct 1080p output. For context on Google's own range, Veo 3.1 Standard is listed on the Gemini API pricing page at $0.40 per second for 720p and 1080p video with audio, so a second of 720p from Omni 1.2 Flash costs 17.4% of the Veo 3.1 Standard price.
··········
SPEED, LATENCY AND QUOTAS: PUBLISHED FIGURES FOR BOTH OMNI MODELS.
Google has published no generation-time figure for either version. The Gemini API guide for Omni 1.1 Flash says generation time varies with duration, resolution and current API load, and the only quantitative speed statement in Google's material is the August claim that 360p generation is up to 60% faster than 720p on Omni 1.1 Flash, measured on system throughput. The Omni 1.2 Flash announcement describes the model as fast without a number.
Rate limits are not stated on any page fetched for this report. The Gemini API model page for Omni 1.1 Flash links to a separate rate-limits page, and Google Cloud lists Provisioned Throughput for Omni 1.2 Flash, which reserves capacity for enterprise customers, without stating its price for Omni on the pricing page read for this report. Teams planning high volumes should confirm quotas in their own Google Cloud project before committing to a schedule.
··········
SAFETY, WATERMARKING AND REGIONAL RESTRICTIONS ON GEMINI OMNI FLASH.
Both models embed an imperceptible SynthID watermark in every output, and the Gemini app lets users upload a file to check whether Google AI generated it. The Gemini API guide for Omni 1.1 Flash says content filters apply to prompts and outputs and vary by region.
The regional limits documented for Omni 1.1 Flash in the Gemini API are specific. Editing and extending uploaded videos is unavailable in the EEA, Switzerland and the United Kingdom; uploading or editing images of minors is not supported in those regions; uploading or editing images of certain recognizable people is not supported anywhere; and YouTube videos cannot be used as a media source. For Omni 1.2 Flash, Google's announcement adds no new safety classification, and the Gemini app help page, which now covers 1.2, repeats the EEA, Switzerland and United Kingdom restriction on video uploads and adds some US states. A Google DeepMind model card page exists for Omni 1.2 Flash, but the page could not be retrieved for this report, so its safety evaluation content is not reflected here.
··········
WHAT GOOGLE HAS NOT PUBLISHED ABOUT OMNI 1.2 FLASH VS 1.1 FLASH.
Several facts a buyer would want are missing from Google's official pages as of November 9, 2026.
........
Item | Omni 1.1 Flash | Omni 1.2 Flash |
Quantitative quality benchmark from Google | None published | None published |
Independent arena rating | Artificial Analysis Elo 1127 | Not rated |
Retirement or shutdown date | None published | Not applicable |
Generation latency | Not published | Not published |
Rate limits on fetched pages | Not stated | Not stated |
Extension limits | 10-second steps, 40 seconds total | Not stated in the announcement |
Gemini for business status | Not announced | Two official posts disagree |
........
The absence of a retirement date for Omni 1.1 Flash leaves developers free to stay on it for now, at a 46% higher video output price per token than 1.2 ($17.50 against $12.00). The absence of any quality score for 1.2 means the improvements in reference adherence, lip-sync and consistency rest on Google's own examples and on customer testimony that Google selected.
··········
GEMINI OMNI 1.2 FLASH OR 1.1 FLASH: WHICH TO CHOOSE BY WORKLOAD.
The decision depends on whether a team values price and clip length today or prefers to wait for independent quality data on the new version. The launch article on Omni 1.2 Flash pricing covers the announcement itself.
........
Workload or user | Recommended model | Reason from the published facts |
High-volume product or catalog video | Omni 1.2 Flash | 31.4% lower video output rate at every resolution |
Draft-heavy creative iteration | Omni 1.2 Flash | Cheaper 360p drafts, same draft-then-upscale workflow |
11 to 15 second ads or scenes in one take | Omni 1.2 Flash | 15-second single generation; 1.1 stops at 10 seconds |
Dialogue with several speakers | Omni 1.2 Flash | Multi-speaker conditioning and audio references (allowlist) |
Enterprise reserved capacity | Omni 1.2 Flash | Provisioned Throughput offered on Google Cloud |
Validated pipeline built on 1.1 behavior | Omni 1.1 Flash, for now | Stable model with no announced retirement and an independent Elo rating |
Clips longer than 15 seconds | Either, with care | 1.1 documents extension to 40 seconds; 1.2 extension rules not stated in the announcement |
Gemini app users | Omni 1.2 Flash by default | The app now runs 1.2; Ultra gets 15-second clips |
........
On the evidence available on November 9, 2026, Omni 1.2 Flash is the version to build on for new projects: it costs $0.0695 per second of 720p against $0.1014, renders 15 seconds where 1.1 renders 10, and is the model Google now ships in its own apps. A team already running Omni 1.1 Flash in production can keep it while it tests 1.2 on its own reference images and scripts, since that test, priced at Google's rates, costs less than the same run on the older model.
··········
FOLLOW US FOR MORE.
·····
DATA STUDIOS
·····
datastudios.org
