Higgsfield Tutorial: Images, Video, AI Influencers and Credits

Higgsfield is a browser-based AI studio that puts its own image models and character tools next to outside models such as Kling, Seedance, Veo and Nano Banana, all paid for from one credit balance. This tutorial takes you from a first image to a first video and a consistent AI influencer, and shows how credits work so you don't burn through a plan in a weekend.

What is inside Higgsfield

The home page can feel crowded the first time. Most of it falls into six tools, according to Higgsfield's help center.

Two more tools come later: Canvas, a visual board that chains models into one workflow, and Supercomputer, an agent that plans and generates a whole project from a brief. You can also drive Higgsfield from Claude through MCP (a standard way to plug tools into AI assistants), which the guide to MCP explains.

Plans and credits in plain English

Higgsfield sells individual plans and team plans. These were the prices on the Higgsfield pricing page in October 2026, before tax; promotions change them often.

A credit is Higgsfield's unit of payment. Each generation costs a set number of credits depending on the model, the resolution and the length, which is why video drains a plan much faster than images.

Mid and top plans also include Unlimited access to selected models for a set period. Those generations don't use credits, but only on higgsfield.ai itself: anything made through MCP, the CLI, Canvas or Supercomputer always costs credits, as Higgsfield's credits guide explains.

Part 1: Your first image with Soul 2.0

Soul is Higgsfield's own image model family. Soul 2.0 is the version to learn first, because it combines a prompt, a style and a trained character in one generation.

1. Open Soul 2.0. Go to Image, then Soul 2.0.
2. Pick a style. Choose a preset, or build a Moodboard from 20 or more of your own reference images in one consistent look.
3. Write the prompt. Subject, setting, light and camera, in one or two sentences.
4. Set the format. Use 9:16 for Reels and TikTok, quality 1.5k or 2k, and a batch of 1 to 4 images.
5. Generate. Results appear in that section and in Assets, your library of everything you have made.

If a prompt feels hard to write, ask Claude to draft it in the order the model reads best:

Two details save frustration. When you attach a reference image in a Soul model, the prompt field switches off and the image leads. And Soul HEX, the Color Transfer option, copies a color palette from any picture you upload.

Part 2: Turn the image into a video

Video costs far more than images, so treat the first takes as cheap drafts. This is the loop that keeps the bill down.

Go to Video and pick a model. Starter includes the Fast and Mini versions of Seedance 2.0, while full Seedance 2.0 and Seedance 2.5 need Plus or higher; Seedance 2.5 makes clips of up to 30 seconds with sound in the same pass.

Kling is the choice for multi-shot scenes. You give it a Start Frame and an End Frame for each scene, pick a genre and a speed ramp, and its Motion Control can copy a movement you record yourself onto your character.

To keep a character identical across separate clips, save it as an Element and tag it with @ in the prompt:

@[YOUR ELEMENT NAME] [ONE CLEAR ACTION] in [SETTING], [LIGHT].
Camera: [ONE MOVE SUCH AS DOLLY IN, TRUCK LEFT, ARC SHOT OR CRANE UP].
Mood: [ONE WORD]. SFX: [TWO REAL SOUNDS IN THE SCENE].

For the wider workflow around clips, from shot lists to editing, read how to make AI videos.

Part 3: How to create an AI influencer

An AI influencer is a virtual character with a fixed face and style that appears across many posts. On Higgsfield you design it, train it once, then reuse it everywhere.

1. Design the character. Open AI Influencer and choose Create new. Set character type, gender, origin, skin color, eye color, skin conditions and age, then refine Face, Body and Style under Advanced settings. Characters render at up to 4K.
2. Make a portrait set. Generate the character from different angles and with different expressions, including at least one full-length shot.
3. Train a Soul ID. Go to Soul, then Train new character, and upload the portraits. Training takes a few minutes.
4. Create content. Pick the character in Soul 2.0 or Soul Cinema for stills. Trained characters also show up as Elements you can tag in Seedance video.
5. Make it move and talk. Use the video button in AI Influencer, record your own movement for Motion Control, and add speech in Lipsync Studio.

If you train a Soul ID on a real person rather than an invented one, use only your own photos or photos of someone who gave permission. For realistic people in full scenes, I also get more natural skin from GPT Image than from Nano Banana Pro, which tends to look glossy in my tests.

When to move up to the studios

Once single clips feel easy, these tools take on bigger jobs. Each one still spends credits from the same balance.

For product images specifically, the AI product photography guide shows how to keep your real product accurate inside these tools.

When a generation fails or looks wrong

• Failed generations. Credits usually come back automatically within minutes. Check Manage Account, then Usage, to confirm.
• Blocked as protected content. Remove anything that could be a real face or copyrighted material from your prompt and references, then retry.
• Stuck in the queue. Draft on Turbo or Fast variants at busy times. Higgsfield asks you to contact support if a generation is stuck for more than 10 hours.
• The face drifts. Retrain the Soul ID on a cleaner set: varied angles, good light, recent photos and one consistent look.
• The camera ignores you. Swap vague words for specific moves such as dolly in, pull back wide, arc right or crane up.

What to do next

Open Higgsfield and make one Soul 2.0 image of your product or character, then animate it at 720p with Seedance and watch the credit counter as you go. Once that loop feels easy, design a character and train a Soul ID, because a consistent face is what turns one-off clips into a series. If you want every part of this demonstrated on screen, the Creators course has a full module on Higgsfield, from the image tools to dealing with failed generations.