Gemini Omni Flash 1.2 in Claude

Gemini Omni Flash 1.2 is the Wireflow node for Google's newest Omni video model, which Google calls Gemini Omni 1.2 Flash. It makes clips of 3 to 15 seconds with the sound written in the same pass, from a text prompt, a start frame, reference images or a clip you want changed.

Claude doesn't render video itself, so the Wireflow connector does that part. Once you add it, Claude plans the shot in your chat and builds a board around the node, and when you ask, it tells you the credit cost and runs the board only after you say yes. The clip comes back as a link, and the board stays in your account for the next version. It's the same connector behind our AI video generation MCP page, pointed at the 1.2 node.

We built and ran the board this page links to, so you can open it and look at real output. It has four nodes and a note.

• The Start Frame Prompt node describes the still, a sage-green teapot and a white cup on a walnut tray in morning window light.
• The Start Frame node runs GPT Image 2.5 Flare at medium quality in 16:9, so you can approve the look before the clip.
• The Shot Prompt node describes the motion and the sound, where a hand pours tea into the cup during a slow push-in, and the trickle, a clink and birdsong are asked for in words.
• The Tea Pour Clip node runs Gemini Omni Flash 1.2 and animates the start frame from the shot prompt into an 8 second 720p clip in a 16:9 frame.

Our run came back as an 8 second 720p clip with a stereo sound track, and the two steps used 94 credits between them. Every value Claude can change is on the node's settings page.

You add the connector once. Custom connectors work in Claude, Cowork and Claude Desktop on Claude's Free, Pro, Max, Team and Enterprise plans.

1. In Claude, open Customize, then Connectors, click + Add and pick Add custom connector.
2. Name it Wireflow and paste https://www.wireflow.ai/api/mcp as the remote MCP server URL.
3. Click Add, sign in with your Wireflow account and approve the permissions. You don't paste an API key anywhere.

Then turn Wireflow on for the chat and ask in plain words. This prompt builds a board like the one on this page for a scene of your own.

I'd like you to use Wireflow to plan an 8 second 16:9 shot of an espresso machine pulling a shot on a cafe counter, with the sound it should have. Can you write a start frame prompt and a shot prompt, build a draft board with a GPT Image 2.5 Flare start frame feeding Gemini Omni Flash 1.2 at 720p, and show me both prompts? Then tell me the credit cost and wait for my yes before you run it.

With that last line, Claude looks up the price and waits for your yes before it spends any credits. It reads the price with list_models, saves the draft with create_workflow, starts it with run_workflow and gets the clip link from get_execution, and Wireflow's guide for Claude walks through each one.

The node's estimate for the default 8 second clip at 720p is 89 credits. Wireflow estimates up to about 167 credits for a 15 second clip, and the price per second drops a little as clips get longer. Higher resolutions cost more as well, and Claude can read the live price for your settings with list_models first, at 0 credits. Our run of this board used 94 credits with the start frame included. Our Gemini Omni Flash 1.2 pricing guide puts those credits next to Google's own price per second.

You can open and change this workflow for free, and new accounts get a free video. More videos need a plan, from $36 a month for 2,660 credits, and every plan comes with API and MCP access.

Because the board stays in your account, Claude can rewrite the shot prompt, set any length from 3 to 15 seconds or switch to 9:16 and run it again, while the earlier take stays on the node.

• Changing values on a board needs the opt-in permission to create, edit and delete workflows, which you tick on the consent screen, and the connector permissions list says what each one allows.
• To change a clip you already have, Claude wires it into the node's video input, up to 15 seconds long, and describes the change, such as new light or a new background.
• check_video looks for frozen regions, silent audio, and black or flash frames before anyone watches the clip, and it costs no credits.
• swap_model puts Veo 3.1 or Seedance 2.5 in the same spot with a price for each, so you can compare models on one prompt.
• Several clips become one video in a Video Editor node on the board, and its exports use no credits.

Google also serves the model through the Gemini API, Google AI Studio, Google Cloud and its own apps such as Flow and Google Vids. If you already pay Google and want a raw API call on your own key, or you make your clips inside those apps, we'd suggest you stay there. Google's own API also takes more reference images per prompt than the 6 the Wireflow node accepts, and the node has no switch that turns the sound off.

Google says the model still has trouble with fast sports action, quick script changes and complex physics, so plan around those shots. Claude and the node give you one shot, and a finished ad still needs a cut, which you can make in a Video Editor node on the board or with the video generator in your browser.