How to Make an AI Video With Higgsfield (Step-by-Step, 2026)

This tutorial follows Higgsfield’s official help center (“How do I create my first generation?”) — the exact screens and button names, in order. No account tricks, no invented steps.

Before you start

• A Higgsfield account (the Free tier works for learning the workflow; most models need a paid plan — see our pricing guide).
• An idea of what you want: text-to-video (describe it in words) or image-to-video (start from a picture).
• Know your credit budget: every generation spends credits (Starter = 270/month). Check the cost tables in our pricing guide before rendering.

Step 1 — Choose “Video” in the top navigation

After logging in, the top navigation splits creation into “Image” and “Video”. Click “Video” — this is the primary mode split the docs use. (The plan comparison table separately groups models under “Text-to-Video” and “Image-to-Video”.)

Step 2 — Select a model or feature

Each section shows the available models. Pick based on your goal, or start with the default the page suggests. Models include Higgsfield’s own (Soul, Soul ID, Genjutsu, Popcorn, Soul Cinema) and third-party options (Seedance 2.0, Kling 3.0, Google Veo 3, Sora 2, Wan 2.6, Pika 2.2, Pixverse V5, Vidu Q2, and others — availability depends on your plan). Model choice is the biggest factor in both quality and credit cost.

Step 3 — Enter your prompt

Describe what you want to create. Be specific about subject, action, camera, and mood — vague prompts produce generic clips. Example structure: “[subject] + [action] + [camera move] + [lighting/mood]”. Keep your first attempts short (5 seconds) — they cost fewer credits while you learn what works.

Step 4 — Upload a reference image (if the model supports it)

For image-to-video, upload your starting frame. Supported models accept a reference image to control composition and subject. Use a clear, high-contrast image — the output follows it closely.

Step 5 — Adjust settings: duration, aspect ratio, resolution

Set duration (shorter = cheaper), aspect ratio (16:9 for YouTube, 9:16 for Reels/TikTok/Shorts), and resolution. Resolution is the biggest cost lever: on Seedance 2.0, 4K costs roughly 9x the 720p Fast version (~110 vs ~12 credits per 5 seconds). Start at 720p for tests; render finals at higher resolution.

Step 6 — Click “Generate”

Click the “Generate” button. Rendering takes from seconds to minutes depending on the model and queue. Do not close the tab mid-render.

Step 7 — Find your video in “Assets”

Outputs appear in the section where you generated them, and every finished generation is collected in “Assets” (“Your generation appears in Assets when it’s done”, per the help center). Download or share from there.

Prompt examples that work

Structure beats poetry. Three templates that work across most models:

• Cinematic clip: [subject] + [action] + [camera move] + [lighting]. Example: “a barista pouring latte art, slow dolly-in, warm morning light.”
• Product shot: [product] on [surface], [background], camera static, studio lighting. Example: “matte black headphones on a concrete table, dark background, camera static, studio lighting.”
• Image-to-video: upload a clean starting frame, then describe only the motion. Example: “camera slowly orbits left, steam rises.”

Keep first attempts at 5 seconds and 720p — the cheapest way to test whether a prompt works before spending more.

Tips that save credits

• Test at 720p and 5 seconds; only render finals at high resolution.
• Higgsfield’s own models are far cheaper than frontier third-party ones (Soul 2.0 images: 0.12 credits vs Nano Banana Pro: 2) — use them for experiments.
• Write better prompts instead of re-rolling: each regeneration spends credits again.
• Track spending per model for a week before choosing a plan — the pricing tables are here.

Common problems

“Model not included on my plan”

The Free tier excludes most models (“Not included” in the plan table). Upgrade or switch to an included model.

“Credits ran out mid-project”

Subscription credits reset monthly with no rollover. Either wait for reset, buy a credit pack (expires in 90 days), or upgrade.

“My video looks nothing like the prompt”

Add camera, lighting, and action specifics to the prompt. If the subject drifts, switch to image-to-video with a reference frame.

Frequently asked questions