Claude Just Became a Video Studio

Claude + Higgsfield does it all from one conversation - here's the setup.

May 26, 2026

I don’t own a camera.

I don’t have a production team.

I have Claude, a Higgsfield account, and 20 minutes.

Last week, I generated four cinematic video clips: all without leaving Claude chat. No drag-and-drop editor. No API keys. Just text.

That’s what the Higgsfield MCP connector does. And by the end of this issue, you’ll know exactly how to set it up and what to say to make it work.

Your gift today: a copy-paste Claude + Higgsfield prompt kit with 6 ready-to-use video prompts, delivered at the end of this issue.

This newsletter grows from your shares. If this saved your time today, send it to one person who’s still generating videos the old way.
Share

What Higgsfield Actually Is

Most people think of Higgsfield as another AI video tool. It’s not. It’s closer to a production studio that you access through a browser.

Higgsfield gives you access to multiple leading AI video models in one workspace:

Seedance 2.0
Kling 3.0
Veo 3.1
Wan
Sora 2 and many others

You can switch between models without leaving the platform, compare outputs side by side, and pick the best result for your project.

That alone separates it from every other tool in this space. You’re not locked into one model. You run the same brief through three engines and keep the winner.

How Claude Talks to Higgsfield

Here’s where it gets interesting.

Higgsfield AI released its MCP server on April 30, 2026, becoming the first platform to bring cinematic-grade image and video generation directly into Claude conversations.

The old way of making AI video looked like this:

Write a prompt → Copy to Runway → Download Clip → Upload to Capcut → Realize the style is wrong → Go Back → Start Over.

The new way:

Describe your video to Claude → Claude picks the model → Video arrives in Chat.

MCP is the bridge. It stands for Model Context Protocol, an open standard that Anthropic developed, which lets Claude securely connect to external tools. Before MCP, connecting an AI to an external API required custom integration work every time.

You’d write wrapper code, handle authentication, manage errors, and maintain the connection yourself. With MCP, a tool like Higgsfield publishes a server that exposes its capabilities in a format Claude already understands.

Once the two are connected, Claude can reach every model and feature on the Higgsfield platform from inside any conversation.

If this is useful, share it with one person who’s been trying to figure out AI video. Takes 10 seconds.
Share

Step-by-Step Guide: Claude + Higgsfield

Step 1: Create a Higgsfield Account

Go to higgsfield.ai
Create an account.

Step 2: Open Claude and Go to Settings

You need Claude on desktop or in the browser at claude.ai.

Open claude.ai
Click your profile icon at the Bottom right corner.
Navigate to Settings → Connectors → Customize → Click on “+” on Top

Step 3: Add Higgsfield as a Custom Connector

This is the one step most people get wrong.

Click “Add custom connector.”
Name it: Higgsfield
Now go to the Higgsfield website.
Click on MCP & CLI.
Now select the MCP toggle.
Copy the URL of “Add a Custom Connector.”
Paste the MCP server URL in the Custom Connector URL.
Click Save

Name it Higgsfield and paste the URL exactly. This line does all the work. If you mistype the URL, the connector will appear to save, but Claude won’t be able to reach Higgsfield’s models.

Step 4: Connect Your Higgsfield Account

Click the Connect that shows in Connectors Heggsfield Custom.
It will ask for your Identity and Email Address Verification. Just allow it.
Next, it will ask for a few permissions. Just Click on Always Allow.
Return to Claude, the connector is now live

You only do this once. After that, every conversation can access Higgsfield instantly.

Step 5: Check Higgsfield Is Connected to Claude

Before generating anything, confirm the connection is actually live.

Open a new Claude Chat.
Click on the “+” icon.
Then click on “connector”; it should show Higgsfield as active.
Then type exactly this in the new chat:

What Higgsfield tools do you have access to right now?

Claude should respond with a list of available models and capabilities, Seedance, Kling, Veo, Soul, and others

If Claude responds with “I don’t have access to any tools,” go back to Step 3 and confirm the MCP URL is pasted correctly. One wrong character breaks the whole connection.

Step 6: Describe Your Video and Let Claude Work

Describe your video the way you’d describe it to a director.

Tell Claude the scene, the camera move, and the vibe.
Specify aspect ratio (9:16 for vertical, 16:9 for horizontal, 1:1 for square).
Mention the model if you have a preference, or let Claude choose.
Wait 30–60 seconds for the clip to render and appear in chat.

Here I will use the pasted prompt to generate an ad for the Supplement Bottle. You can also copy this prompt to create a UGC ad.

I want to generate a UGC-style video ad. Here are the details:

Scene: A woman in her late 20s sits at her desk at home, morning light 
coming through the window behind her. She picks up a black matte 
supplement bottle, looks directly at the camera, and holds it up 
slightly — like she's about to tell a friend why she loves it.

No scripted speech. Just the action. Natural, unposed, authentic.

Shot: Handheld iPhone-style footage. Slight natural shake. 
Close-medium framing — face and product both visible.

Lighting: Warm natural window light. No ring light. No studio setup.

Model: Soul 2.0 for maximum photorealism.
Duration: 5–6 seconds.
Aspect ratio: 9:16 for TikTok and Instagram Reels.
Style: Raw, native, creator-shot. Should look like organic content — 
not a brand ad.

After generating, tell me the model used and one thing I can 
adjust to make it look even more native.

I Tried It. Here’s What Happened

I tried this system for generating a video.

The results were extraordinary with character consistency, motion, sound, Lighting accuracy, and scene transitions.

What This Combo Can Actually Build

This is not a list of fantasy use cases. These are the four workflows I’ve tested personally.

Social Content at Scale: One brief. Multiple platform formats. You describe a scene once and ask Claude to generate it in 9:16 for TikTok, 16:9 for YouTube, and 1:1 for Instagram, all in the same conversation. A single conversation can expand a prompt into hundreds of campaign-ready videos, sized for every platform you publish on.
UGC-Style Product Ads: A Higgsfield-generated UGC ad doesn’t look like a polished corporate video; it’s designed to look like something a real creator posted. This matters because native-looking content outperforms produced-looking content on every short-form platform right now. You describe a real person using your product, specify the Soul model for photorealism, and Claude delivers something that looks like an iPhone-shot testimonial. I tested this for a mock skincare product. The output was indistinguishable from what a real creator would post.
Cinematic Scene Building: Direction translates from plain language into actual cinematographic parameters: camera moves, lens choice, depth of field, aspect ratio, and frame rate. Motion brushes and physics-aware simulation handle action sequences, and time remapping covers slow motion. You’re not just generating video. You’re directing it. Ask for a crash zoom, a dolly push, and an overhead sweep, and Claude translates it into the right parameters automatically.
Recurring Characters Across Videos: You train a Soul Character once from a handful of reference photos and then reuse it across every scene, week after week, with identity holding stable through every render. Voices clone the same way, complete with multilingual lip sync, and you direct emotion, gesture, and wardrobe per take in plain language. This is the capability most tools still can’t pull off. Character consistency, the same face, same style, across 10 different clips, is what turns a one-off video into a series.

This newsletter grows from your shares. If you got value from this, pass it on to one person who needs it.
Share

Your Free Gift - 6 Copy-Paste Prompts

Cinematic Scene Prompt

Use this when you want a high-quality short clip with real camera direction.

Generate a 5-second cinematic video of a misty mountain valley at golden hour.
Camera: slow dolly push forward, starting wide and ending at mid-shot.
Model: Veo 3.1 or Kling 3.0 — you choose based on quality.
Aspect ratio: 16:9.
Style: photorealistic, natural light, no color grading.

Slow dolly push forward, starting wide and ending at mid-shot, this line does all the work. Without the camera direction, Claude generates static or randomized movement. With it, you get intentional cinematography.

UGC Ad Prompt

Use this for social content that looks native, not produced.

Generate a 6-second UGC-style video of a young woman in her 20s applying
a face serum in her bathroom. Shot feels like iPhone footage.
Lighting: natural window light, slightly warm.
Camera: handheld, minimal movement.
Model: Soul 2.0 for photorealism.
Aspect ratio: 9:16 for TikTok.
No text overlays. No branding. Just the action.

Shot feels like iPhone footage, the single phrase that shifts Higgsfield from polished to native. Remove it and you get stock footage. Keep it and you get content.

Multi-Model Comparison Prompt

Use this when you’re not sure which model fits your scene.

I want to generate a 5-second video of a futuristic city skyline at night
with neon reflections in rain puddles.
Run this through Sora 2 and Kling 3.0 separately.
Aspect ratio: 16:9.
Deliver both clips and tell me which one has better motion quality.

Asking Claude to compare models in a single prompt saves 10 minutes of manual testing. This is the workflow, not a trick.

Product Reveal Prompt

Use this for e-commerce or brand content.

Generate a 4-second product reveal video for a minimal black leather wallet.
Shot: object on a dark slate surface, dramatic overhead key light.
Camera: slow push in from wide to close-up, ending on the monogram detail.
Model: Cinema Studio.
Aspect ratio: 1:1 for Instagram.
Style: luxury, clean, no motion blur.

Character Consistency Prompt (After Training a Soul Character)

Using my saved Soul Character [Character Name], generate a 5-second clip
of them walking down a Tokyo street at night.
Expression: confident, slight smile.
Wardrobe: black oversized jacket, white sneakers.
Camera: tracking shot from the front, slight wide angle.
Model: Soul Cinema.
Aspect ratio: 9:16.

Using my saved Soul Character, this is the phrase that unlocks identity consistency. Without it, Claude generates a new person every time. With it, your character holds across every clip you produce.

Batch Campaign Prompt

I'm running a campaign for a new running shoe.
Generate 3 separate clips:
1. Athlete sprinting on a misty track, 9:16, Seedance 2.0
2. Close-up of shoe detail while in motion, 1:1, Kling 3.0
3. Lifestyle shot — person stretching at sunrise in a park, 16:9, Veo 3.1

All clips should feel cohesive: warm light, natural palette, no voiceover.
Deliver all 3 in sequence.

Here are a few issues that you might have missed (go check them out):

If this issue helped you, forward it to one person who needs it. It's free, and it takes 10 seconds.

Sharing is always caring:)

Become a paid subscriber: Get every past issue of AI In Public, plus exclusive resources, prompts, and early access to everything I build next.

And for shorter takes between issues, follow me on X → Hamza Khalid

Hamza 💙

AI in public

Discussion about this post

Ready for more?