Gemini Omni 1.1 Flash: Pricing, Video Quality, and Early Glitches
Google's Gemini Omni 1.1 Flash video model costs 10 cents per 720p clip and tops LMArena's text-to-video leaderboard. Here's a hands-on look.
What is Gemini Omni 1.1 Flash?
Gemini Omni 1.1 Flash is Google’s updated video generation model, released as part of a broader wave of Google announcements. It replaces (or at least overshadows) the company’s earlier VO line, since most of Google’s recent video tooling now runs under the Omni name. The model generates short video clips from text prompts, image references, and start/end frame pairs, and it’s priced per generation rather than per subscription tier.
Pricing sits at 10 cents for a 720p output, matching the cost of the previous Omni model. Google also offers a 360p option for quick test generations, plus 1080p and 4K tiers for when you want a finished, higher-resolution clip. The workflow that makes sense for most users: generate cheap at 360p to check whether a prompt and composition work, then re-run or upscale to a higher resolution once you’re happy with the result.
What can Gemini Omni 1.1 Flash actually do?
The feature set carries over from the original Omni model, with a few upgrades layered on top. According to Google’s announcement, the new version can:
- Analyze up to 10 seconds of prior context, which matters for continuity between generations or extending an existing clip.
- Accept specified start and end frames, letting you animate a transition between two images rather than generating from scratch.
- Upscale output to 4K.
- Take a video reference as input, not just text or a still image.
One coffee. One working app.
You bring the idea. Remy manages the project.
None of this is conceptually new for the category. Start/end frame control and video-to-video referencing have shown up in competing tools already. What’s notable here is that Google bundled these into one model at a flat per-generation price, rather than gating them behind separate products or higher subscription tiers.
How do you access Gemini Omni 1.1 Flash?
There are two paths. Developers can build against it in Google AI Studio using an API key, paying per generation. Consumers who subscribe to Google’s AI Plus, Pro, or Ultra plans can use it inside Google Flow, Google’s video creation interface, without touching the API directly.
In hands-on testing, the AI Studio route produced repeated internal server errors, even after funding an account with API credits. That’s a rough edge worth flagging for anyone planning to build on this model day one: a fresh release from a major lab can still have unstable API endpoints in its first days, and it’s worth budgeting for retries or fallback logic if you’re integrating it into a product pipeline. Switching to Flow (available through an Ultra subscription) got generations working without issue, which suggests the problem was specific to the API layer rather than the model itself.
How does it perform on LMArena?
On LMArena’s text-to-video leaderboard, a blind-comparison system where users are shown two anonymous outputs from different models and vote on which they prefer, Gemini Omni 1.1 Flash currently ranks first. In the image-to-video category, it sits in second place, close behind Minimax, though LMArena’s own vote counts are still low for this model since it’s newly added. Early leaderboard position is a useful signal but not a settled verdict: rankings can shift meaningfully once thousands more votes come in.
That said, a first-place finish in text-to-video blind testing, even an early one, indicates the model is landing well with users when compared head-to-head against other video generators.
What are the quality issues people are running into?
Leaderboard rank and hands-on impression don’t always match, and that’s true here. A test prompt asking for a robot wearing a hat labeled “future tools” breakdancing in front of a castle with a dragon flying overhead produced a clip with real problems: the robot’s head detaches and falls to the ground mid-animation, the hat text renders identically on both sides of the head in a way that doesn’t match the geometry, and the face is inconsistent when viewed from different angles. The castle and background elements held up fine. The failure was concentrated in the character animation and object permanence.
This lines up with a broader pattern in AI video generation right now: backgrounds, environments, and static or slow-moving elements tend to look convincing, while articulated characters doing complex physical motion (dancing, fighting, detailed hand or facial work) are where models still fall apart. A single test prompt isn’t a full benchmark, but it’s a useful reminder that leaderboard wins reflect aggregate voter preference across many prompts, not flawless output on every generation.
Is Gemini Omni 1.1 Flash worth using right now?
- ✕a coding agent
- ✕no-code
- ✕vibe coding
- ✕a faster Cursor
The one that tells the coding agents what to build.
For quick, cheap iteration, yes. A 10-cent 360p test generation is inexpensive enough to experiment freely with prompts before committing to a full-resolution render, and the 4K upscale path means you’re not stuck with a low-res final product. The feature set (start/end frames, video reference, extended context) covers most of what people actually want from a video model in production workflows.
For anything requiring precise, glitch-free character animation, the current hands-on impression suggests some caution. Wonky head detachment, mismatched details across frames, and general inconsistency in articulated motion are the kind of flaws that show up in a spot-check and would need testing across your own specific use case before relying on it for finished work. If your use case leans toward atmospheric or environmental video (the marine biology and microscopic-style examples Google demonstrated), quality looks stronger than for prompts involving detailed character choreography.
Frequently Asked Questions
How much does Gemini Omni 1.1 Flash cost per video?
A 720p generation costs 10 cents. Google also offers 360p for cheaper test runs and 1080p and 4K tiers for higher-resolution final output, with pricing scaling accordingly by resolution.
Where can I use Gemini Omni 1.1 Flash?
Through Google AI Studio via an API key for developers, or through Google Flow if you’re subscribed to AI Plus, Pro, or Ultra. Flow is the consumer-facing interface and doesn’t require direct API access.
Is Gemini Omni 1.1 Flash better than Google’s previous VO video model?
It’s priced the same as the previous Omni model and adds capabilities like extended prior-context analysis and start/end frame control. It’s effectively positioned as VO’s successor, with Google’s recent releases centering on the Omni name rather than VO.
How does Gemini Omni 1.1 Flash rank against other video models?
On LMArena’s blind text-to-video leaderboard, it currently ranks first. In image-to-video, it ranks second, close behind Minimax, though vote totals are still relatively low since the model is newly listed.
What are the biggest weaknesses of Gemini Omni 1.1 Flash?
Hands-on testing surfaced animation glitches in character-focused prompts, including a detaching head and inconsistent facial and hat details across a single clip. Background and environmental generation appeared more stable than articulated character motion.
