Higgsfield Soul 2.0 is an image-generation model. Soul ID is the reusable character identity layer. Keeping those two jobs separate helps you plan a campaign: decide who appears first, then direct the image, style and palette around that identity.
This independent Clout guide explains the official documentation checked October 3, 2026 and adds an original six-shot campaign plan. The downloadable artwork is conceptual reference imagery, not a Higgsfield result or a training dataset. We did not train a Soul ID or benchmark image consistency.
What is Higgsfield Soul and where does Soul ID fit
The official Soul overview separates Soul for preset-led images, Soul 2.0 for more creative controls, and Soul Cinema for cinematic imagery. Soul ID supplies a trained identity; a Moodboard directs style; Soul HEX directs color. They solve different parts of the brief.
| Layer | Question it answers | Campaign example |
|---|---|---|
| Image model | What generates this still | Select Soul 2 for the planned portrait |
| Character identity | Who appears | One original creator across the series |
| Style direction | What the image should feel like | Quiet daylight editorial photography |
| Color direction | Which palette supports the campaign | Teal wardrobe and warm ivory backgrounds |
| Scene brief | What happens in this particular image | A creator reviewing portrait prints |
Searches such as Soul v2 and Higgsfield Soul 2 often refer to the Soul 2.0 model. Check the exact label in your selected interface. A model name alone does not establish access to identity training, video generation or an API operation.
For video, treat the accepted still as a production input rather than assuming an image model also supplies motion or speech. Use the separate Seedance R2V guide when you need to assign character, prop and movement references to a new scene.
Prepare a character identity before the campaign
The Soul ID training guide specifies 20 or more photos of one person, up to 80, with varied angles and expressions and at least one full-height photo. It places training under Soul → Train new character. Use your own likeness or one you have permission to use.
For an invented person, the same help page says to generate their portraits before training. A single attractive hero image is not the required portrait set. Review the proposed character across multiple views before expanding it: otherwise you may collect photographs of several subtly different faces and call them one identity.
Write a short identity brief with observable details. Our fictional example is an adult woman with dark skin, short natural coiled hair and small gold hoops. The teal jacket belongs to this campaign's wardrobe, not an immutable definition of her identity. That distinction helps you assess a later outfit change without treating it as a new character.
Make a contact sheet of candidate portraits for your own review. Compare face shape, hairline, relative feature proportions and visible distinguishing details. Remove duplicates, heavy occlusion and images that appear to show a different person. More files are useful only when they add clear evidence of the same identity.
Separate identity references from a style Moodboard
The official Soul overview describes Moodboards as a Soul 2.0 feature and recommends a cohesive set of 20 or more style images without faces. Soul HEX extracts a palette from an image and is available in Soul 2.0 and Soul Cinema. A style board and an identity dataset should not be interchangeable folders.
For this campaign, put lighting, table materials and restrained teal-and-ivory color references in the style folder. Keep face portraits in the identity folder. The separation lets you replace a background aesthetic without silently changing the character evidence.
A good style brief also states what to avoid: glossy plastic skin, busy props, unreadable packaging or a competing face in the background. These are production preferences, not claims about a model's default behavior. Review the final download against them rather than treating the preset name as an approval.
Choose prompt mode or reference image mode
Current Soul help says attaching a reference image makes the prompt field unavailable; identity and palette controls can still apply. Check the active mode before composing a long prompt. The text briefs below are for a mode with text input or for planning your reference image, not instructions to paste into a disabled field.
With text input, describe one shot with a clear subject, action, location and camera framing. With a reference image, select a composition that actually expresses those decisions. Do not attach a distant full-body fashion photograph and expect an unrelated tabletop composition simply because your planning notes describe it.
Start with an uncomplicated baseline. Review a neutral portrait before asking for unusual perspective, dramatic light or an extreme facial expression. If the baseline looks wrong, adding more styling makes diagnosis harder. Save the approved image and its settings so a later variation has a concrete comparison.
Plan six images around one recognizable creator
Our original example is a small editorial campaign about a creator building her personal brand. The outputs have different jobs; six nearly identical portraits would not cover the same needs.
| Asset | Scene direction | Acceptance check |
|---|---|---|
| Profile portrait | Neutral ivory background with clear face | Identity and avatar crop |
| Workday story | Creator reviewing portrait prints | Hands, print edges and scene clarity |
| Wardrobe detail | Teal jacket texture in a closer framing | Garment color and fabric appearance |
| Location variation | Same creator at a quiet studio window | Face under different light |
| Campaign cover | Wider composition with usable negative space | Title placement after export |
| Video starting still | Simple seated pose with visible hands | Composition suitable for the intended motion |
Choose the required crop for each asset before generation. A portrait that works as a profile thumbnail may hide the wardrobe detail needed by the campaign. Leave planned space for your actual headline, then add typography in an editor where spelling and placement remain controllable.
Keep the same identity baseline while changing one scene decision at a time. Once the location variation is approved, you can test a different crop against it. Changing the face reference, lighting, outfit and camera angle together gives you no useful explanation for a failed image.
Download the original planning artwork and shot briefs
Download the original full-resolution campaign artwork. It illustrates a fictional creator comparing printed portraits. It is one composed scene, not twenty separate training photos and not evidence that Soul produced matching faces.
Profile portrait brief
Use the approved original creator identity. Adult woman with short natural coiled hair and small gold hoops, teal tailored jacket over an ivory top. Head-and-shoulders portrait against a warm ivory studio wall. Soft daylight, relaxed neutral expression, realistic skin texture, face fully visible. Leave room around the hair and shoulders for a square crop. No text or logos.
Workday story brief
Use the same approved creator and campaign wardrobe. She sits at a pale oak table and reviews one portrait print. Keep the hands clearly visible and the print separate from her fingers. Quiet daylight editorial scene with a terracotta vase and a folded teal fabric swatch. Medium-wide framing, natural expression, no added lettering or interface elements.
Campaign cover brief
Use the same approved creator, teal jacket and ivory top. Place her beside a studio window on the right side of a wide composition. Reserve a clean ivory wall area on the left for a headline added after export. Calm daylight, recognizable face, restrained background, no extra people, no generated text or watermark.
Download the shot briefs and six-image plan. They are original, untested creative directions. Use the input mode supported by your provider and revise from the output you actually observe.
Fix character drift without losing the useful baseline
The Soul ID help page acknowledges that results are not pixel-identical and that weak datasets, extreme angles or style changes can cause drift. It also says the trained identity stays inside the Higgsfield ecosystem rather than exporting as a standalone file. Downloaded images are separate assets.
Review a changed image next to the baseline at a useful size. If only the jacket color changed, inspect palette and wardrobe direction. If facial proportions changed, inspect the selected identity and the source set. If the hand merges with a print, simplify the physical interaction. These are different failures with different next actions.
Keep rejected attempts out of the approved character folder. Naming a file approved because it was generated successfully blurs the difference between completed processing and usable campaign content. Your review record should say what failed and what you changed next.
Download the blank campaign review worksheet. It lists the six planned assets but leaves the observation, cost, time and decision fields empty. Fill those from actual work; no success rate or generation speed is prefilled.
Choose a creator workflow for your next campaign
For your own reusable narration, use the Higgsfield clone guide with original recording text and a separate audio review plan.
Check the selected operation and displayed price in your current account before generating. This guide does not promise a universal free Soul allowance or a particular subscription entitlement. The Higgsfield free-access guide and Unlimited guide address those separate questions.
If your main goal is a recurring original character with coordinated photos and video scenes, build that campaign in Clout. We publish this guide and include Clout for this creator use case, without claiming it exports a Soul ID or matches another model's output. Continue with the consistent creator workflow and AI lookbook guide, or review Clout pricing for your campaign.
