Guide2026/06/01Updated 2026/07/06

Gemini Omni Flash API Status: What's Live in Public Preview (and What Isn't)

Gemini Omni Flash hit public preview on June 30, 2026 — model ID and price are live. What shipped, what preview means, and how it compares to Veo 3.1.

Gemini Omni Flash API Status: What's Live in Public Preview (and What Isn't)

Quick Answer

The Gemini Omni Flash API is now in public preview. On June 30, 2026, Google opened the model to developers through Google AI Studio and the Gemini API under the model ID gemini-omni-flash-preview, at $0.10 per second of output — the six-week wait behind "coming weeks" is over. Public preview is not general availability: the model string carries a -preview suffix, quotas are conservative, and Google can change parameters before GA. This post is the status-and-history companion to our hands-on API tutorial — here you get the timeline, what "public preview" actually means for a production roadmap, the launch feature set, and how Omni Flash stacks up against Seedance 2.0, Kling 3.0, and Veo 3.1. For the model ID, parameter table, pricing math, and runnable Python, JavaScript, and curl, go straight to the tutorial. Consumer access via the Gemini app, Google Flow, and YouTube is unchanged; what's new is that you can now call the model in code.

Key Takeaways

  • Public preview since June 30, 2026 via Google AI Studio, the Gemini API, and the Gemini Enterprise Agent Platform — not GA, so expect the surface to change before it stabilizes.
  • Model ID gemini-omni-flash-preview, $0.10 per second of output — a 10-second clip is about $1.00, matching Veo 3.1 Fast.
  • "Coming weeks" took six weeks. Announced May 19, consumer-only for the first stretch, developer API on June 30.
  • 720p, 3–10-second clips. No 1080p/4K in the preview; Veo 3.1 stays the API pick for higher resolutions.
  • Conversational editing is the real differentiator — the workflow, not raw pixel quality, is where Omni leads.
  • Want the code? The model ID, parameters, pricing table, and working examples live in the API tutorial. This post covers status, history, and positioning.

The status on July 6, 2026

Here is the precise state of play, stripped of hype:

  • Announced: May 19, 2026, at Google I/O.
  • Consumer rollout: Immediate and global, to Google AI Plus, Pro, and Ultra subscribers via the Gemini app and Google Flow, plus no-cost generation in YouTube Shorts and the YouTube Create app.
  • Developer/enterprise API: Public preview since June 30, 2026, through Google AI Studio, the Gemini API, and the Gemini Enterprise Agent Platform.
  • Model ID: gemini-omni-flash-preview, published in the Gemini API docs.
  • Pricing: $0.10 per second of output video. A 10-second clip costs about $1.00.
  • Output: 3–10-second clips at 720p. No 1080p or 4K in the preview.
  • Avatar mode: Still held back. The ability to record your own voice and likeness was demoed but is not part of the public preview while Google tests "how we can bring this capability to users responsibly."

If you searched "Gemini Omni Flash API" for a model string and a price-per-second, both now exist — gemini-omni-flash-preview at $0.10/sec. For the setup steps and runnable code, see the Gemini Omni Flash API tutorial.

What "public preview" actually means

Public preview is a real, callable API — but it is not general availability, and the difference matters for anything you're shipping to users:

  • The model string carries a -preview suffix, Google's standard signal that parameters, quotas, and behavior can change before GA.
  • Quotas are conservative. Preview-tier rate limits aren't published on the pricing page; check your AI Studio project's quota panel before load-testing.
  • Breaking changes are fair game. A parameter rename or a default change can land without a long deprecation window.

The practical takeaway hasn't inverted so much as softened: you can build on Omni Flash today, but isolate the call behind your own interface so a preview change doesn't ripple through your product, and don't hard-depend on preview quotas for a launch with a fixed ship date.

The road from "coming weeks" to public preview

For six weeks after the May 19 debut, Omni Flash was consumer-only and Google's line never budged from "coming weeks" — no model ID, no waitlist, a blank pricing row. That's the window this post originally documented, skeptically. On June 30, 2026, Google's launch post closed it: Omni Flash entered public preview in AI Studio and the Gemini API, alongside the generally available Nano Banana 2 Lite image model. "Coming weeks" turned out to mean about six.

The lesson holds for the next Google model launch: consumer access can ship months before a callable API, and "coming weeks" is a range, not a date. Treat a published model ID and price — not a keynote — as the signal to integrate.

What shipped: the launch feature set

  • 10-second clips, synchronized audio, conversational editing, SynthID watermark — the core capability set, now callable.
  • Text, image, and short-video inputs. Video references are capped at 3 seconds in the preview.
  • Conversational editing via the Interactions API. You chain requests to refine a clip across turns instead of re-rolling it — the mechanics are in the tutorial and the conversational editing guide.
  • Avatar mode held back for safety review; not part of the public preview.

How Omni Flash actually stacks up

Enough time has passed for the early-adopter dust to settle, and the picture is more nuanced than the launch-day demos suggested. Here's the honest read.

On raw generation quality, Omni Flash is not the leader. On the Artificial Analysis Video Arena leaderboard, ByteDance's Seedance 2.0 sits first in both text-to-video and image-to-video, ahead of Kling 3.0, Veo 3.1, and the now-deprecated Sora 2. Multiple creator assessments place Omni's pure pixel quality behind Seedance and Kling. If you only care about the single most photoreal 5-second clip, Omni is not automatically your pick — and the preview caps you at 720p.

On workflow, Omni Flash leads. Where Omni wins is the thing the leaderboards don't score: conversational editing and multimodal input in a single generation call. You describe a change in plain language — "warm the light, pull the camera back" — and the previous frames anchor the edit instead of you rewriting a full prompt and rolling the dice. For iterative, multi-shot work, that's a categorical advantage. We covered the mechanics in Conversational video editing with Gemini Omni Flash.

So the verdict is split by job:

  • Single hero shot, maximum fidelity: Seedance 2.0 or Kling 3.0 today.
  • Iterative, directed, multi-shot work: Omni Flash's editing loop, now that you can call it.
  • Narrative scenes with native audio, 1080p, and 4K: Veo 3.1 remains the production all-rounder.

This isn't a knock on Omni — it's a young, preview-stage model whose differentiator is the workflow, not the benchmark. Pick your tool for the job in front of you, not the keynote.

Ready to build? Start with the tutorial

If you want to wire Omni Flash into a product, the companion post has everything this one deliberately leaves out: the model ID, the parameter table, the pricing math, the current limits, and working Python, JavaScript, and curl for text-to-video, image-to-video, and multi-turn editing.

And if you'd rather not write integration code against a preview API at all, an independent platform like Omni Flash gives you an Omni-era generation and conversational-editing workflow in the browser — prompt, generate, refine — with no API key, SDK setup, or preview churn to manage.

FAQ

Is the Gemini Omni Flash API available right now?

Yes — in public preview since June 30, 2026. Google opened the model to developers through Google AI Studio and the Gemini API under the model ID gemini-omni-flash-preview, at $0.10 per second of output. It is not generally available: the -preview suffix means Google can change parameters, quotas, and behavior before GA. Consumer access through the Gemini app, Google Flow, and YouTube remains live too.

Is the Gemini Omni Flash API generally available (GA)?

No. It is in public preview, which is a real, callable API but not GA. Expect conservative quotas and the possibility of breaking changes before Google promotes it to general availability. For production work that needs GA stability today, Veo 3.1 remains the conservative pick.

How much does the Gemini Omni Flash API cost?

$0.10 per second of output video, so a 10-second clip is about $1.00 — the same rate as Veo 3.1 Fast. Conversational-edit turns bill at the same per-second rate because each turn renders new output. The full cost breakdown and a per-length table are in the API tutorial.

What is the Gemini Omni Flash model ID?

gemini-omni-flash-preview. It runs on the Interactions API rather than Veo's generate_videos pattern. See the tutorial for setup and runnable examples.

Can I use Gemini Omni Flash without the API?

Yes. It's live in the Gemini app and Google Flow for AI Plus, Pro, and Ultra subscribers, and free inside YouTube Shorts and the YouTube Create app. That's enough to prototype, learn the model, and run real production work without touching code.

Is Omni Flash better than Seedance 2.0 or Kling 3.0?

It depends on the job. On raw video quality, Seedance 2.0 and Kling 3.0 currently rank ahead on the Artificial Analysis leaderboard. Omni Flash's edge is its conversational editing and multimodal workflow, which the benchmarks don't measure. For a single high-fidelity shot, reach for Seedance or Kling; for iterative multi-shot direction, Omni's editing loop is the differentiator. See our Gemini Omni Flash alternatives roundup for the full field.

What is avatar mode, and why is it held back?

Avatar mode lets you record your own voice and likeness to generate videos that look and sound like you. Google demoed it but withheld it from the public preview while it tests how to release the capability responsibly — a caution rooted in deepfake and consent concerns. It is not part of the current preview.

Should I build on Sora 2 instead?

No — Sora is shutting down. OpenAI discontinued the Sora app and web experience on April 26, 2026, and the Sora 2 API shuts down on September 24, 2026. Building a new integration on a deprecated model with a hard end-of-life date isn't worth it. See Gemini Omni Flash vs Sora 2 for the comparison.

Is Omni Flash affiliated with Google?

No. Omni Flash is an independent platform built for the Gemini Omni era. It is not affiliated with Google. Gemini, Gemini Omni, Omni Flash, Veo, Sora, Seedance, Kling, and related names are trademarks of their respective owners.

Next steps