How to Use Higgsfield AI: A Step-by-Step Guide

Higgsfield is a subscription studio that brings 15-plus third-party video and image models, including Veo 3.1, Kling 3.0, and Seedance 2.0, under one account. It also includes ad-focused tools such as Product-to-Video and Soul, which help keep a brand consistent across shots. For a complete overview, see our Higgsfield AI for business explainer.

This guide focuses on the practical part: how to use Higgsfield, step by step, from creating an account to exporting a finished clip. You don't need any prior experience with the platform.

Because Higgsfield updates regularly, screens and menu names may change. Treat the steps below as a guide to the overall workflow rather than a pixel-perfect walkthrough, and check the current interface at higgsfield.ai before you begin.

Step 1: Create an Account and Pick a Plan

Sign up at higgsfield.ai with an email or a social login. New accounts may get a small amount of trial credits to test model quality (Higgsfield is not truly free for ongoing work, so treat any trial credits as a demo, not a working tier) — see our Higgsfield pricing guide for the current, honest answer on what "free" actually gets you.

Most people end up picking a paid tier at or shortly after signup, since steady creation burns through trial credits quickly. That guide also breaks down the Starter, Plus, and Ultra tiers and what each one's credit pool actually buys — worth reading before you pick a tier, since the right plan depends on your monthly output, not the feature list.

• Trial credits (if offered): enough to judge model quality, not for ongoing output.
• Paid tiers (Starter/Plus/Ultra): the working tiers for anyone publishing regularly.
• Business tier: per-seat pricing if more than one person on your team needs access.

Step 2: Choose Your Starting Point — a Product Photo, a Prompt, or an Existing Clip

Higgsfield is not one model, it is a control panel over many. Your first real decision is what you are feeding it: a product photo (for Product-to-Video or Draw-to-Video), a text prompt (for straight text-to-video generation), or an existing image you want animated.

For product and ad work — the use case Higgsfield is built around — start with a clean product photo. The clearer the source image (plain background, good lighting, product fully visible), the fewer re-rolls you will need once you get to generation.

• Product-to-Video: upload a static product photo, get an animated ad clip with automatic camera moves.
• Draw-to-Video / Higgsfield Ads 2.0: place a specific product, outfit, or object directly into a generated scene.
• Straight text-to-video: skip the product photo and generate from a written prompt instead.

Step 3: Pick the Underlying Model for the Job

Every generation runs on one of the bundled engines, and picking the right one is the single biggest lever on both quality and credit cost. Higgsfield includes Google Veo 3.1, Kling 3.0, Seedance 2.0, WAN 2.6, and Hailuo 02 for video, plus Flux Kontext, GPT Image, and Nano Banana Pro for stills. Sora 2 was available in Higgsfield until OpenAI closed the Sora API on September 24, 2026.

Match the model to the job rather than always reaching for the most expensive one. Cheaper engines like Kling and Seedance are built for fast drafts and cost far fewer credits per clip; reserve Veo 3.1 for the final, client-facing cut once you know the shot works.

• Draft first on a cheaper model (Kling, Seedance) to lock the composition and motion.
• Re-run the approved draft on a premium model (Veo 3.1) only for the final export.
• For product placement specifically, use Draw-to-Video or Ads 2.0 rather than a generic text-to-video model.

Step 4: Generate, Review, and Re-roll

Submit the generation and expect to iterate. Consistency is the hard part of AI product video — a product's label, color, or logo can shift subtly between takes — so budget for more than one attempt on any shot you plan to actually use.

If you are keeping a product or character consistent across multiple shots, use Soul or Soul Inpaint to hold that identity steady rather than regenerating each clip independently and hoping they match.

• Expect to re-roll: plan credits for 2-3 attempts on any hero shot, not just one.
• Use Soul / Soul Inpaint when a product or character needs to look identical across a sequence of clips.
• Pick a small number of hero angles rather than a long, complex multi-shot sequence — drift compounds with shot count.

Creating the Imagined Zombie Transformation by Hand

Higgsfield does not document a dedicated one-click preset for the imagined zombie transformation trend. Check the effects and presets inside Higgsfield first, and build the sequence by hand if none matches the trend. Start from a clear image-to-video source photo of someone who agreed to it, and keep the prompt non-graphic.

• Split the sequence into two separate shots, generating the real person first and the imagined zombie version second to avoid asking one clip to morph.
• Use Soul or Soul Inpaint across both generations to help hold the person's identity steady between the real shot and the zombie shot.
• Draft motion tests on a cheaper model such as Kling or Seedance, and only re-run the approved take on Veo 3.1 for the final version.
• Plan for two to three attempts per hero shot, and keep the total clip count low because visual drift compounds across sequential generations.
• Join the two finished clips in any video editor.

Step 5: Export and Reuse

Once a generation clears review, export it from your project and drop it into whatever you are publishing to — an ad platform, a social scheduler, or a landing page. Unwatermarked export requires a paid tier; the free tier's output carries a watermark.

Keep the source product photo and your chosen model/settings noted somewhere outside Higgsfield. If you need a same-product variant later (a new angle, a seasonal update), starting from the same source image gets you back to a consistent look faster than starting from scratch.

Content Guidelines and Moderation Flags in Higgsfield

Higgsfield bars sexually explicit, unlawful, and defamatory content, and requires you to confirm permission for all visible faces in an uploaded image. Uploading faces without consent can lead to account suspension. We could not read the full list of prohibited categories, so read the complete rules directly in the Higgsfield Terms of Use agreement. Higgsfield updated its terms in July 2026, and the current terms are in force on that page.

Moderation works across two distinct layers. Each integrated model applies its own automated filters to prompts, reference images, and outputs. In addition, public Community content passes through automated screening filters alongside a user-report system. Higgsfield reviews flagged content and removes posts that violate guidelines, while serious violations can lead to account suspension or termination.

When a generation gets blocked as not safe for work (NSFW), the flag may come from an external model provider. Higgsfield has no access to the internal moderation rules of external model providers. Higgsfield support can help rephrase prompts to reduce false positives, as explained on the Higgsfield NSFW troubleshooting guide, but bypassing or circumventing moderation filters is prohibited.