Gemini Omni 1.2 Flash Pricing and How to Use the API
15-Second Video at About $0.07/sec (720p)
Gemini Omni 1.2 Flash went GA Nov 5, 2026: 3-15s video at about $0.07/s (720p), ~30% below 1.1. Per-resolution prices, monthly costs, API code, limits, rivals
Gemini Omni 1.2 Flash is Google's video generation and editing model, made generally available on November 5, 2026. Clips can now run 3 to 15 seconds in one-second steps, up from a 10-second maximum on Omni 1.1 Flash, and 720p costs about $0.07 per second, roughly 30% less than Omni 1.1. The model ID is gemini-omni-1.2-flash, available through the Gemini API (Google AI Studio) and Google Cloud (Gemini for business).
The prices below come from the official pricing page: about $0.023/s at 360p, about $0.07/s at 720p, $0.104/s at 1080p and $0.209/s at 4K (1080p and 4K are upscaled). A 15-second 720p clip therefore costs about $1.05. There is no free tier and no batch pricing listed. This article reflects information as of November 2026 and prices may change.
Three-line summary
- Length: 3 to 15 seconds in one-second steps (Omni 1.1 topped out at 10). The API default is 10 seconds
- Price: about $0.07/s at 720p (about $1.05 for 15 seconds), $0.104/s at 1080p, $0.209/s at 4K. No free tier
- Migration: from December 3, 2026, requests to gemini-omni-1.1-flash are routed to gemini-omni-1.2-flash and billed at 1.2 rates
What changed from Omni 1.1 Flash
| Item | Omni 1.1 Flash | Omni 1.2 Flash |
|---|---|---|
| Model ID | gemini-omni-1.1-flash | gemini-omni-1.2-flash |
| Clip length | Up to 10 s | 3 to 15 s (1 s steps) |
| Video output price (per 1M tokens) | $17.50 (about $0.10/s) | $12.00 (about $0.07/s at 720p) |
| Per-resolution pricing | Not listed on the page | 360p / 720p / 1080p / 4K listed |
| Conversational editing | Supported | Works across full 15 s clips (official blog) |
| Going forward | Auto-routed to 1.2 from Dec 3 | Current version |
According to Google's blog, 1.2 keeps characters, products and scenes more consistent from first frame to last, with fewer jump cuts and steadier scale in complex camera moves. Dialogue, lip-sync and multi-speaker conditioning are also said to be better. These are Google's own claims, not independent tests. For the previous generation, see our Gemini 3.5 Flash and Gemini Omni overview.
Pricing
On the Gemini API pricing page, video is billed as output tokens, with one second of 720p video counted as 5,792 tokens. The output price is $12.00 per 1M tokens (1.1 Flash: $17.50). 5,792 tokens x $12.00 / 1M is about $0.0695 per second, which is where the $0.07 figure comes from.
| Resolution | Per second | Per 15 s (calculated) | Note |
|---|---|---|---|
| 360p | about $0.023 | about $0.35 | Drafts and framing checks |
| 720p (default) | about $0.07 | about $1.05 | Matches the $0.07 in the blog |
| 1080p | $0.104 | about $1.56 | Upscaled |
| 4K | $0.209 | about $3.14 | Upscaled |
Omni 1.1 Flash is listed at $17.50 per 1M video output tokens, which the page describes as roughly $0.10 per second, with no per-resolution breakdown. Against 720p at $0.07/s that is about 30% cheaper. Both models show the free tier as not available, and no batch pricing is listed.
Monthly cost estimates (with the math)
These are our own estimates from the unit prices above, at $1 = 150 yen, excluding edits and extra turns (we could not confirm from the official pages how edits are billed).
| Usage | Calculation | Monthly |
|---|---|---|
| 100 clips/month, 720p, 15 s, one shot | 100 x 15 s x $0.07 = $105 | about $105 (about 15,700 yen) |
| 3 drafts at 360p, then 720p final (15 s) | 100 x (3 x $0.345 + $1.05) = about $208 | about $208 (about 31,300 yen) |
| 30 clips/month, 1080p, 15 s | 30 x 15 s x $0.104 = $46.8 | about $47 (about 7,000 yen) |
| 10 clips/month, 4K, 15 s | 10 x 15 s x $0.209 = $31.4 | about $31 (about 4,700 yen) |
Cost scales with the number of attempts. Locking in composition at 360p and rendering only approved clips at 720p or above keeps failed attempts cheap compared with running everything at 720p.
Differences by channel
| Channel | Role | What we could confirm |
|---|---|---|
| Gemini API / Google AI Studio | Developer API, GA | Model ID gemini-omni-1.2-flash. No free tier. Docs say Provisioned Throughput is not supported |
| Google Cloud (Vertex AI) / Gemini for business | Enterprise, GA | The blog lists Provisioned Throughput. We could not confirm the Vertex AI price table |
| Flow (Google Labs) | Multi-turn direction and edits | Stated in the official blog |
| Google Vids | 15-second clips | No additional cost with a Google account or Workspace subscription (official blog) |
| Gemini app | 15-second videos | Rolling out to Google AI Ultra subscribers (official blog) |
| Google Ads | Video campaign creation | Stated in the official blog |
One caveat: the Omni page in the Gemini API docs says Provisioned Throughput is not supported, while the official blog describes it for Google Cloud. If you need dedicated capacity, check the latest Google Cloud documentation.
How to use the API
In the documented Python SDK example, you call client.interactions.create with the model ID and a prompt, then decode and save the returned video data. This is the official form.
import base64
from google import genai
client = genai.Client()
interaction = client.interactions.create(
model="gemini-omni-1.2-flash",
input="A marble rolling fast on a chain reaction style track, continuous smooth shot."
)
with open("marble.mp4", "wb") as f:
f.write(base64.b64decode(interaction.output_video.data))Length, resolution and aspect ratio are set by parameters.
"duration": "15s", # Supported values: "3s" to "15s" (default: "10s")
"resolution": "1080p", # 360p, 720p (default), 1080p (upscaled), 4k (upscaled)
"aspect_ratio": "9:16" # "9:16" or "16:9" (default 16:9)If you request URI delivery (response_format={"type": "video", "delivery": "uri"}), wait until the file state is ACTIVE before downloading.
# response_format={"type": "video", "delivery": "uri"} polling pattern when you set(as shown in the official docs)
while True:
f_info = client.files.get(name=f"files/{file_name}")
if f_info.state.name == "ACTIVE":
break
elif f_info.state.name == "FAILED":
raise RuntimeError("Generation failed.")
time.sleep(5)
client.files.download(file=video_output.uri, destination="output.mp4")What it can do, and the limits
- Length: 3 to 15 seconds. Longer values return an invalid argument error. Extension uses the last 10 seconds of the input and adds 3 to 10 seconds per turn, up to 30 seconds in total
- Resolution: 360p / 720p (default) / 1080p and 4K (both upscaled). Aspect ratios 9:16 and 16:9
- Inputs: text, images and video. Up to 6 reference images (up to 3 per character, 2 characters). Reference videos: 1 clip for 11 to 15 s outputs, up to 3 clips (3 s each) for 3 to 10 s outputs
- Editing: conversational edits to generated or uploaded videos of 15 s or less. With store=false, later turns cannot edit the result
- Audio: per the docs, audio inputs and audio references are unsupported in the current API, audio in reference videos is ignored, and voice editing is not supported. Videos themselves are generated with audio
- Other limits: only English is fully supported; no system instructions, temperature, top_p or negative prompts; YouTube videos cannot be used as input; first/last-frame interpolation covers 3 to 10 s only
- Watermark: generated videos carry an invisible SynthID watermark that can be detected programmatically (official docs)
Google's blog names fast sports action, rapid script changes and complex physics as areas still being improved, so expect weakness there. Commercial-use terms and rights to outputs should be checked in the terms of service and the generative AI additional terms; we do not draw a conclusion here.
Comparison with competitors
Only products with an officially confirmed per-second price are listed. OpenAI's Sora was not found on its official pricing page and Kling's official page could not be read directly, so both are left out. The Runway rows are prices through the Runway API (1 credit = $0.01, Runway's documentation), which differ from each vendor's direct pricing.
| Service / model | Price per second | Note |
|---|---|---|
| Gemini Omni 1.2 Flash (Google) | about $0.07 (720p) / $0.104 (1080p) / $0.209 (4K) | Up to 15 s. Google's official pricing page |
| Runway Gen-4.5 | $0.12 (12 credits) | Runway API |
| Runway Gen-4 Turbo | $0.05 (5 credits) | Runway API |
| Veo 3.1 (via Runway API) | $0.40 with audio / $0.20 without | Price through Runway API |
| Veo 3.1 Fast (via Runway API) | $0.15 with audio / $0.10 without | Price through Runway API |
| Kling 3.0 Standard (via Runway API) | $0.13 with audio / $0.09 without | Price through Runway API |
| Seedance 2 (via Runway API) | $0.30 (720p) / $0.40 (1080p) | Price through Runway API |
On unit price alone Gen-4 Turbo is cheaper, but maximum length, resolution, audio and editing differ, so no fair ranking is possible. In practice, run the same prompt under comparable conditions (length, audio, resolution) and compare the total cost per finished clip. Our Gemini 3.8 Flash pricing article covers Google's text-model pricing in the same period.
Business use and cautions
- Good fits: short product videos, vertical (9:16) social clips, ad concept drafts, store announcement drafts
- Likeness: avoid videos resembling real people without their permission, and check rights and consent for photos used as references
- Copyright and trademarks: avoid prompts that evoke existing characters or brands, and review output manually before commercial release
- SynthID: the watermark is invisible and helps verify provenance. Any on-screen AI disclosure is a separate decision under platform rules and your own policy
- Quality control: only English is fully supported, so check any speech, subtitles or signage text rather than shipping it as generated
Sources
- Google official blog: Introducing Gemini Omni 1.2 Flash
- Gemini API pricing
- Gemini API video generation docs
- Gemini Omni guide
- Runway API pricing guide
How much does Gemini Omni 1.2 Flash cost?
The official pricing page lists video output at $12.00 per 1M tokens, with one second of 720p counted as 5,792 tokens, so about $0.07 per second. 360p is about $0.023/s, 1080p is $0.104/s and 4K is $0.209/s (1080p and 4K are upscaled). There is no free tier.
How much cheaper is it than Omni 1.1 Flash?
The pricing page lists 1.1 Flash at $17.50 per 1M tokens, about $0.10/s, versus about $0.07/s for 1.2 Flash at 720p. Google's blog calls it roughly a 30% cut.
What is the maximum clip length?
3 to 15 seconds per generation, in one-second steps. The API default is 10 seconds. With the extension feature you can add 3 to 10 seconds per turn, up to 30 seconds in total.
Until when can I use Omni 1.1 Flash?
Per the Gemini API pricing page, from December 3, 2026 requests to gemini-omni-1.1-flash are routed to gemini-omni-1.2-flash and billed at 1.2 rates.
Can I use an audio reference?
The current Gemini API docs say audio inputs and audio references are unsupported, and audio in reference videos is ignored. The blog demo showed short voice references, so check for API updates.
Are generated videos marked as AI-made?
They carry an invisible SynthID watermark that can be detected programmatically. No visible on-screen label is added, so any disclosure when publishing is a separate decision.
Feel free to contact us
Contact Us