Google Gemini Omni Flash: The Complete Guide to Google's Any-to-Any AI Video Model

What Makes Gemini Omni Flash Different: Any-to-Any Multimodal Input

Unlike pipeline models that convert inputs sequentially, Gemini Omni Flash processes text, images, audio, and video as a single unified input and produces video output — a capability called "any-to-any" ^AAB. The model generates 3–10 second 720p clips at 24 fps in 16:9, 9:16, or 1:1 aspect ratios, with synchronized native audio created in the same pass ^HG.

You can feed the model up to seven reference images, one video clip, and a text brief in a single prompt, and it merges them into a coherent output ^BG. For example, a user could provide product photos, a background video, and a description, and Omni Flash would generate a short promotional clip.

Conversational Video Editing: The Interactions API

The headline capability is conversational video editing, enabled by the Interactions API ^AA. Instead of exporting clips to a separate editing tool, users can iteratively refine outputs through natural language — character swaps, product swaps, style transfers, object additions, and relighting — while the model preserves the parts of the video the user wants to keep ^AA. Each turn in the conversation produces a new video, and the model understands context from prior turns ^A.

Pricing: $0.10 Per Second

API pricing is set at $0.10 per second of generated video, equivalent to $1 per 10-second clip ^AH. This matches Veo 3.1 Fast, positioning Omni Flash as a lower-cost, faster-iteration alternative for video workflows ^Y. Pricing is also $1.50 per million input tokens and $17.50 per million video output tokens ^H.

For consumers, Gemini Omni Flash is free on YouTube Shorts for users aged 18 and older ^AWN. It is also included in Google AI Plus, Pro, and Ultra subscription tiers ^AG.

The 10-Second Clip Limit: A Deliberate Choice

The 10-second cap is an intentional product decision at launch, with longer durations planned for future releases ^HY. Google has characterized this as a responsible rollout strategy, balancing capability with safety ^T. Some commentators note the limit may be a deployment-side limitation rather than a model architecture limit ^N.

SynthID Watermarking and Content Provenance

Every clip generated or edited with Omni Flash includes an invisible SynthID digital watermark ^YNN. Users can verify AI-generated content through the Gemini app, Gemini in Chrome, and Google Search by asking, "Was this video created using Google AI?" ^NB. On supported Google surfaces, content also carries C2PA Content Credentials ^NC. There is no API knob to disable it ^DG.

Adobe Firefly Integration

On July 1, 2026, Adobe added Gemini Omni Flash as a partner model within Adobe Firefly, enabling plain-English video editing using Firefly's interface and credit system ^D. It costs 30 credits per second of video ^D. This follows a broader partnership announced at Google I/O to bring the Adobe connector to Gemini ^B. Users can access it from the Model dropdown in the Text to Video module ^A.

Companion Model: Nano Banana 2 Lite

Released alongside Omni Flash, Nano Banana 2 Lite (model ID: gemini-3.1-flash-lite-image) is the fastest and cheapest image model in Google's Nano Banana family ^AA. It generates images in under 4 seconds at under $0.04 per 1,000 images ^BHB. General availability was announced June 30, 2026 on Google Cloud, and it is available in Google AI Studio, the Gemini API, and the Gemini Enterprise Agent Platform ^HCB.

Omni Product Studio

Google showcased Omni Product Studio at Google I/O 2026, a demo app that turns static product images into cinematic e-commerce videos using Gemini Omni Flash ^GC.

Regional Restrictions and Limitations

Editing uploaded videos is blocked for users in the European Economic Area (EEA), Switzerland, and the United Kingdom ^A. Community reports also note blocks in India and some US states ^A. The model also reliably edits only its own generated clips — not arbitrary third-party uploads — as a safety measure against deepfake misuse ^A. Speech and audio editing capabilities are deliberately withheld at launch over deepfake concerns ^TGL.