Gemini Omni 1.1 Flash: settings, prompts and a real example

Gemini Omni 1.1 Flash is Google's audio-and-video model, documented with controls for revising a shot after generating it: drafting at a lower resolution, guiding a transition with start/end frames, extending a scene and preparing a higher-resolution result. Cicadas runs the core text-to-video and image-to-video generation; treat the frame, extension and upscaling controls described later in this guide as Google's own documented capabilities, not steps available on this page.

Try it on Cicadas: Gemini Omni 1.1 Flash runs on the Gemini Omni Flash page. Generate from text or one image, 4, 6 or 8 seconds, 720p or 1080p, 16:9 or 9:16. Every clip includes generated audio—there is no silent option. A 4-second 720p clip is quoted at 100 credits before you generate (200 credits for 8 seconds at 720p, 300 for 8 seconds at 1080p). Failed generations refund credits automatically, and every successful clip is a private MP4 you can download and recover later in My videos.

The recorded example

Generated on September 27, 2026: text-to-video, 4 seconds, 720p.

The prompt asked for a ginger cat asleep on a stack of books by a rainy bookshop window, with soft rain and distant thunder. The visual matched the brief, but automated audio analysis heard soft ambient music rather than the requested rain and thunder. That's a useful reminder for this model: describing a specific sound in your prompt is a request, not a guarantee, so always review the actual audio track before you rely on it.

Build your brief

Choose a shot that can be judged on a few concrete points, and describe the sound you want along with the picture. Here is an untested extra idea, not a second recorded example:

A small red paper boat floats from the left third toward the center of a shallow puddle. The camera stays at ground level. Soft overcast reflections ripple behind it. One continuous movement, no new boats, no text, ending with the boat near the center.

Decide whether a stable boat shape or a precise final position matters more before you generate, and change one thing at a time so the result tells you what improved. Don't ask for a new subject, a complex action and a large camera move in the same first attempt.

What Google documents beyond this page

Google's August 27 announcement describes scene extension, start/end-frame control, 360p drafting and 4K upscaling. These are announced product capabilities, not steps you can run on Cicadas' Gemini Omni Flash page, which currently offers text-to-video and image-to-video only.

The model reference identifies the stable model as gemini-omni-1.1-flash. It lists text, image and video inputs, ordinary output clips of 3–10 seconds, and video input up to 10 seconds for editing or extension—a generated shot's length and the eventual length of an extended sequence are different limits, and extension is not offered on this page. For start/end frames or extension, see Google's Flow controls update directly, or use Cicadas' first and last frame video page, which covers the same two-frame idea with a different model.

Review and finish

Review an ordinary-speed playback before isolating frames, then check faces, logos and straight edges—in our example, the cat's paws and the book spines are a good place to start. Use the frame extractor for close inspection. Keep source frames and accepted clips organized by shot; version names are more useful than a folder full of final files.

For continuity and revision work across several clips, see the product video maker.