- Genial
- AI Automation Tutorials
- Claude + Higgsfield: AI Marketing Assets in Minutes
Claude + Higgsfield: AI Marketing Assets in Minutes
You will connect Higgsfield to Claude as a custom connector, then ask Claude in plain words to generate product images and UGC-style videos with several models at once and pick the best result.
What you need
A Claude account
A Higgsfield account (higgsfield.ai)
A product or brand to showcase, with a short description of its look
Step by step
Copy the Higgsfield MCP server URL
On the Higgsfield homepage, open the MCP tab and copy the server URL shown there. You will paste it into Claude in the next step.
Add Higgsfield as a custom connector in Claude
In Claude, click your profile, open Settings, then Connectors. Scroll to the bottom and click Add Custom Connector. Enter the name Higgsfield, paste the server URL, click Add, then connect your Higgsfield account. You only do this once.
Connector nameHiggsfield
Check that Claude is connected
Start a new chat and ask Claude to confirm the connection. Claude scans the tools it has access to and tells you whether the Higgsfield MCP works. If it fails, the server may simply be down, so try again later.
Connection checkHey, Claude, are you connected to the Higgsfield MCP server?
Ask for two product images with two models
Describe your product and branding, and name the two image models you want. Swap in your own brand details. Claude checks the models are available on Higgsfield, writes the prompts and picks the format (for example 1:1 square for social posts).
Image prompt (adapt the brand details)I've been working a lot on my new brand of protein powder called Berserk Powder. What I want you to do is create two new images, one with GPT Image 2 and the other one with NanoBanana 2, showcasing the product. The branding is dark with red accents.
Review the images and pick one
Compare the two results side by side. From here you can ask Claude to edit an image, animate it, or download it and use it as is. If an image preview does not show in Claude, it may still exist in Higgsfield: this is a known issue with the new connector.
Turn the chosen image into UGC videos
Tell Claude which image you liked and ask for videos from two models at once. Describe the scene you want. Claude writes the detailed video prompt (vertical 9:16, TikTok style) for you.
Video prompt (adapt the product and action)Perfect. Now what I want you to do is I want you to take the NanoBanana 2 image, and I want you to create two different videos of about eight to ten seconds using Kling 3.0 and Seedance 2.0. It must be an UGC format with a guy showing the protein powder and how he dry scoops it.
Double-check model names if you dictate
If you use speech-to-text (I use WisprFlow), check how model names were transcribed. In the video, Seedance 2.0 came out as "C9 2.0". Claude usually works out what you meant, and asks you if it cannot.
Find your assets in Higgsfield while you wait
Videos take about 30 seconds to a minute, longer with the bigger pro versions (the demo took about two minutes). Meanwhile, open the Assets folder in Higgsfield: it lists everything you have generated, so nothing is lost if the Claude chat breaks.
Compare the videos and keep the best
Watch both videos and choose the one you prefer (in the demo, Seedance 2.0 won). You can generate up to four videos in parallel and keep only the ones you want, and create 4K videos when you need them.
Watch out for
MCP connections can fail when the server is down. Ask Claude to verify the connection before you start generating.
A generated image may not preview in Claude even though it was created. Check the Assets folder in Higgsfield.
Speech-to-text tools can mangle model names (Kling, Seedance). Read your message before sending it.
Video generation is slower than images: expect 30 seconds to a few minutes, more with pro model versions.




