Getting Started with Higgsfield

Section 1

What Higgsfield Is (and What It Is Not)

Higgsfield is one subscription that gives you most of the major AI image, video, and audio models in a single workspace, plus tools for keeping characters and products consistent across shots.

Instead of paying separately for Kling, Veo, Sora, Seedance, Nano Banana, GPT Image, and a voice tool, you buy credits on Higgsfield and switch between models from one menu. Everything you make lands in one library, and you can reuse it as the starting point for the next generation.

The three building blocks, plus one

What Higgsfield is not

• Not free to run. Every generation costs credits, and better models, higher resolution, longer clips, and bigger batches cost more.
• Not a full video editor. You will usually finish projects (cuts, music, captions) in a normal editor such as CapCut, Premiere, or DaVinci Resolve.
• Not the easiest interface. Menus hide until you hover over them and contrast is low. Section 3 shows you how to cope, and Section 9 shows how to avoid most of it.
• Not one-and-done. Expect to generate a few versions of most shots. Good results come from iterating cheaply and only finalizing the winner.

Section 2

Choosing a Plan and Signing Up

Higgsfield runs on subscriptions that include a monthly pool of credits. You spend credits each time you generate something.

1. Go to Higgsfield’s pricing page to see the current plans and promotions.
2. Choose between the plan tiers. At the time of writing the main tiers were Starter, Plus, and Ultra; higher tiers add more credits and some extra features. Names and prices change often.
3. Use the toggle to switch between monthly and annual billing. Annual is cheaper per month if you know you will keep using it.
4. Create your account and complete checkout.

Which plan? Everything in this guide works on the entry-level plan. Start there, learn how quickly you use credits, and upgrade only if you run out. If you plan to drive Higgsfield from Claude (Section 9), note that the connector requires an active paid subscription.

Section 3

Finding Your Way Around

Let’s be honest: Higgsfield’s interface is not easy to navigate. It is dark, low in contrast, and many buttons and menus stay hidden until your mouse passes over them. Once you know that, it becomes manageable.

The first screen after you log in is the busy Explore page. Ignore most of it. Almost everything you need is reached through the thin menu bar along the top of the screen, and that menu bar is where most new users get stuck.

Why it feels like things are missing

• Hover menus. The Image, Video, and Audio tabs look like plain words. Their real menus (features on the left, models on the right) only appear when you hover over them. Clicking is not enough on its own; you have to hover and wait for the panel to open.
• Low contrast. Dark gray text and icons on a near-black background are easy to miss, especially on a laptop screen or in a bright room.
• Hidden controls. Options such as the three-dot menu on a generation, Download, and Create Element often appear only when your cursor is over the item.
• Constant change. Higgsfield adds features almost weekly, so items move and new ones appear. Even experienced users have to hunt occasionally.

Tricks that make it easier

• Turn up your screen brightness while you work in Higgsfield.
• Zoom the browser in (Ctrl and + on Windows, Cmd and + on Mac) so text and icons are larger.
• Use a full-size window. On narrow windows some menu items collapse or move.
• Bookmark the pages you use most once you find them, such as Create Image, Create Video, and Cinema Studio, so you can skip the hover menus next time.
• Or skip the interface entirely for most tasks by driving Higgsfield from Claude (Section 9).

What lives where

Section 4

Generating Your First Image

Images are the cheapest way to learn Higgsfield, and they make better starting points for video than text alone.

Hover over Image in the top menu and wait for the panel to open; then move straight down into it and click Create Image. Past generations fill the middle of the page; the bar along the bottom is where you set everything up. Working left to right:

Write the prompt, then generate

Describe three things: the subject (what is in the picture), the setting (where it is), and the style (lighting, mood, camera feel). For example:

A matte black cold brew can with the attached logo, on a weathered wooden cafe counter, morning window light, shallow depth of field, editorial product photography.

Then click Generate. When it finishes, click the result to open it and use Download to save it.

The same pattern works for almost anything: upload a selfie and ask for a superhero portrait, or upload your product mockup and ask to see it “in real life” in a store or a customer’s hand. You can also skip the reference entirely and describe something from scratch.

Section 5

Generating Your First Video

Video works best when you start from an image, because the model already knows what the scene looks like.

Hover over Video, wait for the panel, and click Create Video. Controls sit on the left; finished videos appear on the right. Three tabs run across the top:

Step by step

1. Pick a model. The General option used Kling 3.0 by default in recent tests; it works on every plan and can generate sound. Newer models (such as Kling 4.0 and Seedance 2.5) appear in the model list as they launch.
2. Set a Start Frame. Click it and choose an image, such as the product shot from Section 4. An optional End Frame tells the model where the clip should finish.
3. Describe the action in the prompt: what moves, how the camera moves, and what you should hear. “The can is cracked open, condensation drips, slow push-in, soft cafe ambience.”
4. Check the three toggles under the prompt: Enhance (Higgsfield polishes your prompt; leave it on), the speaker icon (generate sound with the video), and Elements (attach saved characters, locations, or props).
5. Set duration and resolution, then click Generate. Play the result and use the download button to save it.

Section 6

Generating Audio and Voiceovers

Hover over Audio and wait for its panel to open to find three tools. The layout matches the video page: controls on the left, results on the right.

Creating a voiceover

1. Click Text to Speech.
2. Optionally upload a voice sample if you want it to sound like a specific person (use only voices you have the right to use, such as your own).
3. Type your exact script in the script box.
4. Describe the voice in plain English under voice details: “warm, friendly, mid-30s, calm and upbeat.”
5. Choose the model and batch size, click Generate, play it, and download it.

Section 7

Cinema Studio: Putting It All Together

Cinema Studio is where you direct. It is built for multi-shot projects, such as an ad or a film scene, where the same character, location, and product must look identical in every shot.

Hover over Video and click Cinema Studio. It opens as its own workspace with a Projects list and an Elements library in the sidebar, plus the familiar generation bar with Image and Video tabs.

Step 1: Create a project

Click New Project under Projects, give it a name, and click Create. Everything you generate inside it is saved to the project.

Step 2: Build your elements

An element is a saved character, location, or prop. Make each once, then reuse it in any shot.

Step 3: Direct the shot

1. Click All Assets, then switch the generation bar to the Video tab.
2. Set the Cinema Studio extras or leave them on Auto: Genre (overall feel), Style (color and lighting), and Camera (camera body and lens, as on a real set).
3. Write the prompt using @ to insert elements: “@Maya sits at the counter of @Harbor Cafe, holding a @Ridgeline Can, smiles and takes a sip.” The attached elements appear above the prompt.
4. Set aspect ratio, resolution, duration, batch size, and audio. Generate the first version at 480p to check motion cheaply.
5. When you like the result, open it and choose Edit > Upscale, set 1080p, and click Upscale. The upscaled file appears in Assets, not inside the project.

To leave Cinema Studio, click the Cinema Studio menu at the top and choose Higgsfield AI to return to the main dashboard. Your project stays in the sidebar for next time.

Section 8

Assets, Credits, and Finishing Your Project

Where your work lives, how to watch your credits, and how to turn clips into a finished ad.

• Assets (top right) holds everything you have generated, including upscaled videos and anything made through Claude or other connected agents.
• Profile (top right) shows your plan, remaining credits, and billing. Check it before large batches.
• Finish in an editor. Download your images, clips, and voiceover, then assemble them in CapCut, Premiere, DaVinci Resolve, or similar. Add music, captions, and your logo end card there.

Section 9

Driving Higgsfield from Claude (MCP)

Between the hover menus, low contrast, and dozens of toggles, the web app can be slow going. Higgsfield’s official MCP connector lets you skip most of it: you tell Claude what you want in plain English, and Claude does the clicking.

MCP (Model Context Protocol) is an open standard that lets an AI assistant use outside tools. Higgsfield runs an official, hosted MCP server. Once connected, Claude can generate images, video, and audio in your Higgsfield account, choose an appropriate model, use your saved characters and elements, upscale, and check your credit balance, all from a normal chat. Results are saved to your Higgsfield Assets, just like web generations.

What you need

• An active paid Higgsfield subscription. No API key is needed; you sign in with your Higgsfield account.
• A Claude account on claude.ai or the Claude Desktop app that supports custom connectors.

Connect Higgsfield to Claude (web and desktop)

1. In claude.ai or Claude Desktop, open Settings > Connectors (some versions label this area Customize).
2. Click Add custom connector.
3. Name it Higgsfield and paste the server URL:
   https://mcp.higgsfield.ai/mcpCopy
4. Click Connect. A Higgsfield sign-in window opens; sign in and authorize. If you are already logged in to higgsfield.ai, this happens automatically.
5. Back in Claude, confirm Higgsfield is in your list of connectors and switched on for the conversation.

Check that it works

Start a new conversation and ask: “What is my Higgsfield credit balance?” Claude should call a Higgsfield tool and report your balance. If it answers without using a tool, check that the connector is toggled on for that chat and try again.

Claude Code and terminal agents

For Claude Code, Higgsfield recommends its command-line tool rather than the connector. Paste this into Claude Code and it will set itself up:

Set up Higgsfield for me so I can generate images and videos from here.
1. Install the CLI: npm i -g @higgsfield/cli
2. Authenticate: higgsfield auth login
   (complete the sign-in in the browser window it opens)
3. Install the companion skills: npx skills add higgsfield-ai/skills

If you sign in to Claude Code with the same account you use on claude.ai, connectors you added on claude.ai may also appear there automatically. Run /mcp in Claude Code to check.

Prompts that work well

Using your own images

Claude cannot pass an image you drop into the chat straight to Higgsfield. Instead, say “I want to use a photo from my computer as the reference.” Claude opens a Higgsfield upload window in the conversation; pick the file there. You can also paste a direct image URL, or refer to a past generation, a saved element, or a Soul character by name. If Claude asks you to attach the file to the chat instead, remind it: “Open the Higgsfield upload window for my file.”

What you can do through Claude

Image and video generation with all models, upscaling, background removal, expanding images and reframing video, Kling motion control, Soul characters and reference elements, voiceovers, voice cloning and changing, dubbing, turning YouTube videos into short clips, and checking your balance and past generations. Higgsfield also offers skills, preset multi-step workflows for things like ad campaigns, UGC-style product videos, and faceless social content; see higgsfield.ai/skills.

Troubleshooting

Section 10

Best Practices and Common Mistakes

A few habits save most of your credits and most of your frustration.

Save credits

• Image first, video second. Perfect a still frame cheaply, then animate it.
• Draft low, finish high. Lowest resolution, shortest duration, batch size one, until the motion is right.
• Upscale only the winner. Do not generate at 1080p to find out whether you like something.
• Reuse elements instead of re-describing a character in every prompt.

Get better results

• Be specific about subject, setting, action, camera movement, lighting, and sound.
• One idea per clip. Short clips with a single action are more reliable than complex scenes; stitch them together in your editor.
• Keep Enhance on for video unless you need exact wording.
• Expect a few re-rolls. Character consistency can drift between shots; elements and Soul characters reduce this.

Avoid trouble

Section 11

Quick-Reference Checklist

Tick items off as you go. The PDF version has a printable page.

Getting set up

• Plan chosen via the FrostyStack Higgsfield link; monthly or annual decided
• Practiced hovering over Image, Video, and Audio until their panels open; bookmarked Create Image and Create Video
• Higgsfield connected to Claude: Settings > Connectors > custom connector > https://mcp.higgsfield.ai/mcp
• Tested with “What is my Higgsfield credit balance?”

Every project

• Generate the key still image first (product, character, or scene)
• Save recurring characters, locations, and props as elements
• Animate with the image as the Start Frame, at low resolution first
• Generate the voiceover with Text to Speech
• Upscale only the final clips to 1080p
• Download from Assets and finish in your video editor

When working through Claude

• Ask for the credit cost and approval before every generation
• Name the model when it matters; otherwise let Claude choose
• Use the Higgsfield upload window for your own photos