Is there a Claude Code skill for CGI ads?

Yes, Wireflow's CGI ad generator skill for Claude Code puts your product, giant, in a real street as a 5 second 9:16 video, for about 175 credits.

Last updated View as MarkdownAgent setup

Yes, Wireflow's CGI ad generator skill, cgi-ad-generator, makes CGI ads in Claude Code: your product, giant, in a real street, as one 5 second vertical video with sound. From a product photo and an idea in plain words, Claude plans a fake out-of-home (FOOH) ad, writes the opening frame, the final frame and the action, shows the credit cost (about 175 credits a run) and waits for your yes. A Wireflow app then makes both stills with GPT Image 2.5 Sunburst Edit and a 5 second 9:16 Seedance 2.5 shot between them, and Claude saves all three files to ./wireflow-outputs/.

The planning happens in Claude, and the generation runs as a published Wireflow app over the Wireflow MCP connector, signed in as you. Without the connector, the skill still writes the plan sheet and all three prompts, which work in any tool that edits a photo and makes video between two frames. You can read the skill itself in the cgi-ad-generator folder on GitHub.

How do I install the CGI ad skill in Claude Code?

Install the Wireflow plugin, which adds every Wireflow skill plus the Wireflow connector:

/plugin marketplace add wireflowINC/skills
/plugin install wireflow-skills@wireflow

Claude Code shows the plugin's details first, so you can review it and choose its scope. Then run /mcp, choose wireflow and sign in with your Wireflow account. The skill shows up as /wireflow-skills:cgi-ad-generator, and Claude also picks it up from plain requests.

To add only this skill, use the skills CLI, or copy the skills/cgi-ad-generator folder into ~/.claude/skills/ (all projects) or .claude/skills/ (one project). Then connect Wireflow yourself and run /mcp to sign in. The connector uses OAuth, so there is no API key to copy.

npx skills add wireflowINC/skills --skill cgi-ad-generator
claude mcp add --transport http wireflow https://www.wireflow.ai/api/mcp

What do I tell Claude to make a CGI ad?

Tell Claude three things in plain words: your product photo, where the product goes (a city, a kind of street, a landmark) and the one impossible thing that happens. A half idea is enough to start:

Make a CGI ad of our can, huge, in Times Square.

Claude takes whatever you give and asks, in one message, only for what is missing: for the request above, the photo and what happens. If you only name the product, it offers three ideas that follow the skill's idea rules (the product is the impossible thing, it reads with the sound off, the place adds something, and something happens), and you pick one.

The product photo must be at an https:// URL: a clean shot, front on, with the label readable. A photo on your computer goes up through create_upload_link: you open the link, then get_upload_link returns the https URL.

Claude uses the skill when you ask for a CGI ad, a fake out of home ad, a FOOH ad, a giant product in a city, a viral 3D street ad, or a CGI product video for Instagram or TikTok. For a plain product clip on a clean background, use the ecommerce-product-video skill instead.

How does the skill make a CGI ad look real?

It plans the ad with rules that keep the product from looking like a sticker, starting with a plan sheet for the size, the light and the camera height, because those decide whether the ad looks real. The plan follows these rules:

  • Street-scale products stand 2 to 3 times a person's height.
  • The light is shade or overcast, unless you ask for sun.
  • The camera is a phone at eye level or lower, so the top of a tall product is seen from below.
  • A passer-by, a lamppost or a branch sits in front of the product's lower part.

The app holds these fixed as well: a phone camera at eye level or lower, product colours slightly muted, something in front of the product, shadows matched to the people, and the label kept letter for letter.

From that one plan, Claude writes three prompts and sends them to the app with your photo:

Input What it holds
product_photo The https:// URL of a clean product photo, front on, label readable
opening_frame The first frame: the real place, where the product stands, its height in metres beside a person, the light, the camera height and what stands in front of it
final_frame Only what has changed by the last frame. Everything else stays put
what_happens The action from first frame to last, with times in seconds, then the sounds

Claude reads the three prompts side by side before it quotes. The first moment of what_happens should be what opening_frame shows, its last moment what final_frame shows, and the same people stay in the same places. Then it shows you the idea in one line and the three prompts. All four inputs go with every run, because an empty one falls back to the app's own demo, a giant noodle cup in Naples.

What files do I get from one CGI ad run?

You get three files from one run: two stills and a 5 second vertical video with sound.

Output Format
opening-still PNG, 1152x2048
final-still PNG, 1152x2048
ad-video MP4, 9:16, 720p, 5 s, with sound

The app makes both stills with GPT Image 2.5 Sunburst Edit: the opening still from your product photo, then the final still edited from it. Seedance 2.5 then runs from the first still to the second, so the video opens on the opening still and ends on the final still, with real sound only and no music or voice.

Claude starts one run_app per idea you approve, then polls get_execution every 10 seconds while the status is PENDING, RUNNING or IN_PROGRESS, and keeps you posted. The skill's test run took just under 5 minutes from start to finish.

Claude downloads the files with curl, or Invoke-WebRequest in PowerShell, into a folder for the date, each named after the idea:

./wireflow-outputs/YYYY-MM-DD/cgi-ads/idea-slug-opening.png
./wireflow-outputs/YYYY-MM-DD/cgi-ads/idea-slug-final.png
./wireflow-outputs/YYYY-MM-DD/cgi-ads/idea-slug.mp4

It ends with the three files, a link to each and the credits used, worked out as your balance before minus after. It then offers a second shot from further back to prove the size, or a new idea, quoted the same way, and the skill's board in Wireflow for longer shots, 16:9 or your own sound effects.

What happens if the run fails or the ad comes out wrong?

If the run ends as FAILED, CANCELLED or TIMEOUT, Claude stops and quotes the error, and if the ad comes out wrong, it names every miss plainly and asks you before a retake. Seedance screens prompts and images and can refuse a run after the stills are made, most often for copyright, and Claude passes that message on as it is.

Claude looks at both stills before the video and checks that:

  • the label is spelled right, letter for letter
  • the size is close to the plan next to the people
  • the camera cannot see into the top of the product
  • something stands in front of it, and its shadows point the same way as the people's

If ffmpeg is installed, Claude pulls frames at 4 per second and checks that the product is in shot at 0.0 s, that something changes by 2 s, that the main moment lands by about three quarters of the way, and that the clip plays forwards from the opening still to the final still with the label intact. It cannot hear the sound, so it confirms there is an audio track and asks you to listen.

A retake is a full run, about 175 credits, so Claude asks first and changes one thing at a time.

How many credits does one CGI ad cost?

About 175 credits per run: 20 for each still and 135 for 5 seconds of Seedance 2.5, from Wireflow's price list on 16 November 2026. The skill's own test run was charged 174 credits.

What Credits
Opening still 20
Final still 20
5 seconds of Seedance 2.5 135
One run, all three files About 175
A retake About 175, a full run

The plan, the prompts and the quote happen in Claude, and nothing is spent until you say yes. Before anything runs, Claude calls whoami and get_credit_balance and gives you the total, as in this example from the skill:

1 CGI ad, 2 stills and a 5 second Seedance 2.5 shot with sound: about 175 credits. Balance: 1,240.

For several ideas, it lists each one and adds them up. If your balance is short, it says so and stops. Running the skill needs a Wireflow account with credits: see pricing for what each plan includes.

Can I use the CGI ads commercially?

Yes: Wireflow's Terms of Service say content generated by AI models through Wireflow is yours to use for personal and commercial purposes, subject to the underlying AI model provider's terms.

The skill adds its own guardrails: use products you own or may advertise, never copy another brand's ad, logo or campaign, and never use a real person's face. These ads show events that never happened, so the skill suggests the platform's AI or CGI label and never presents an ad as news or as a real installation.

Wireflow's Privacy Policy also says Wireflow does not use your prompts, images or generated content to train AI models.

Does the CGI ad skill work in Cursor, Codex or Claude Desktop?

Yes, once the agent is connected to the Wireflow MCP server. The skills CLI installs the skill in Cursor, Codex, Gemini CLI and other agents that read SKILL.md, though the skills are built for Claude Code and each skill's README notes that Wireflow has not tested them in other agents itself. Connection steps are on the Cursor and Codex pages, and for other clients in Use Wireflow from your agent.

The skill is also written for Claude. In Claude.ai or Claude Desktop, add https://www.wireflow.ai/api/mcp as a custom connector. Saving the files needs a shell.

To start runs from your own code instead of an agent, see the run API docs. For Wireflow's other skills, see the agent skills page.

Related

  • Skills

    Yes, Wireflow's public agent skills make images, videos, product shots and ads in Claude Code, and the skills CLI installs them in Cursor and Codex.

For AI agents: the full documentation index is at /llms.txt, and most docs pages are available as Markdown by adding .md to the URL.

© 2026 Wireflow. All rights reserved.