Workflow spotlight / prompting guidance

AI influencers: build a persona that stays recognizable

A consistent AI influencer starts with an original identity and a reference workflow, not a different face in every post. Soul 2.0/Soul ID is our editorial starting point; reference-led video then adds motion. No follower growth, revenue or engagement outcome is promised.

Last verified 2026-10-07 · Editorial verdict
Create with Soul 2.0

Affiliate link: we may earn a commission if you subscribe. Your model choice comes first. How links work

Editorial shortlist / in workflow order

The models behind the workflow.

01

Soul 2.0

A consistent synthetic persona workflow. Start with Soul when the job is an identity-led persona rather than a one-off illustration. Soul ID is the companion workflow for retaining a character across sessions; it is not a guarantee of identical output.

See capabilities
02

Nano Banana Pro

Deliberate diagrams and designed imagery. Shortlist Pro for an image that depends on labels, composition or several instructions. Google documents text and diagram improvements; every label still needs proofing.

See capabilities
03

Seedance 2.5

Reference-led stories and selective edits. The strongest capability fit in this shortlist for a reference-heavy brief or a clip longer than 15 seconds. Thirty-second output and editing are documented; consistency still needs human inspection.

See capabilities

Ordering reflects editorial workflow fit, not an original quality benchmark or paid placement.

The reason behind the choice

Why this approach makes sense.

A recurring character can make a visual series easier to recognize. That is editorial reasoning, not a measured growth claim. Current vendor tutorials show active workflow interest; we have not measured search or social growth.

Observed signal / not a growth metric

Vendor workflow coverage and dedicated AI Influencer/Soul ID tools; unquantified interest, not a growth-rate claim.

First-seen date and adoption rate: not verified.

A repeatable path

From brief to output.

  1. Define an original adult fictional persona: age, facial structure, hair, wardrobe and tone. Avoid borrowing a real person’s identity.
  2. Create a neutral identity sheet: front, profile and three-quarter portraits. Keep one coherent visual brief.
  3. Save the identity through Soul ID or retain the same reference set; record the exact model and settings.
  4. Make three still scenes changing only the setting. Inspect face shape, age, hair and body proportions against the sheet.
  5. Animate a selected frame in Seedance 2.5 with one action and restrained camera motion; preserve the identity description.
  6. Catalog-backed motion starting point (untested guidance): Seedance 2.5, mode=omni_reference, duration=5, resolution=720p, aspect_ratio=9:16, bitrate_mode=standard. Put the identity frames in image_references. Preflight the current cost; do not transfer the text-only 35-credit snapshot to this setup.
  7. For a silent scene set generate_audio=false. For a speech test set it true and describe the exact short line; audio support does not establish lip-sync accuracy.
  8. For speech, use a short line, then review mouth movement, timing and factual claims.
  9. Disclose the synthetic persona and any commercial relationship; never fabricate a genuine customer testimonial.
Prompting guidance / untested

A concrete starting brief.

Original adult fictional travel creator with short dark curls, warm brown eyes and a charcoal overshirt. Three-quarter portrait in a small bookshop, soft window light, natural skin texture. Preserve reference facial structure and wardrobe. No text.

Suggested settings: use the aspect ratio required by delivery, the shortest valid duration for the scene, and the exact reference roles exposed in the model picker. No unverified preset or credit estimate is implied.

Original image outputs / 7 October 2026

Inspect the actual image runs.

One generation per model per prompt (n=1). This is a snapshot, not a statistical ranking. Each model ran at its Higgsfield default resolution; GPT Image 2 used medium quality. Native sizes differ. Credits are preflight estimates, not verified billed usage.

Identity caveat: the Nano Banana Pro request returned backend nano_banana_2. Its images are not confirmed Pro outputs; both IDs remain visible.

View the full AURA product shot test, settings and A/B slider →

View the full Ramen shop UGC realism test, settings and A/B slider →

Nano Banana Pro request · Nano Banana 2 backend — product-shot-aura
product shot aura

Nano Banana Pro request · Nano Banana 2 backend

Editorial observations: AURA is legible, with light silver-colored beveled lettering. Blue light rims the case and slab; droplets cover the slate, while the lid is mostly dry. The background falls out of focus.

2752×1536 native · 2 credits (preflight) · 2026-10-07

Prompt and generation provenance

Studio product photo of a matte black wireless earbud case on a wet slate surface, soft blue rim light, tiny water droplets, the word AURA embossed in silver on the lid, shallow depth of field, premium advertising look

Requested: nano_banana_pro
Backend: nano_banana_2
Job: 3e9ef002-4f8b-48be-9cad-bd7895f30826

{"platform":"Higgsfield","aspect_ratio":"16:9","resolution":"2k","quality":"model default","other":"model defaults"}

Requested Nano Banana Pro; Higgsfield reported nano_banana_2. This output is not a confirmed Pro benchmark.

Download full-resolution original · Model context

Create with Nano Banana Pro

Affiliate link: we may earn a commission if you subscribe. Your model choice comes first. How links work

Soul 2.0 — product-shot-aura
product shot aura

Soul 2.0

Editorial observations: AURA is legible but looks recessed and blue-tinted rather than silver embossed. The slate is relatively dry with sparse droplets. A cyan rim runs across the upper edge.

2048×1152 native · 0.12 credits (preflight) · 2026-10-07

Prompt and generation provenance

Studio product photo of a matte black wireless earbud case on a wet slate surface, soft blue rim light, tiny water droplets, the word AURA embossed in silver on the lid, shallow depth of field, premium advertising look

Requested: soul_2
Backend: text2image_soul_v2
Job: 9141e77a-788a-46a0-bded-d065dd1d6d25

{"platform":"Higgsfield","aspect_ratio":"16:9","resolution":"2k","quality":"model default","other":"model defaults"}

Higgsfield-reported backend ID recorded; provider internals not independently verified.

Download full-resolution original · Model context

Create with Soul 2.0

Affiliate link: we may earn a commission if you subscribe. Your model choice comes first. How links work

Nano Banana Pro request · Nano Banana 2 backend — ugc-realism-ramen
ugc realism ramen

Nano Banana Pro request · Nano Banana 2 backend

Editorial observations: The woman laughs above a steaming bowl. A moving raised hand is visibly blurred. OPEN LATE neon appears behind her and reversed in the window; fine skin texture and water specks are visible.

2752×1536 native · 2 credits (preflight) · 2026-10-07

Prompt and generation provenance

Candid smartphone photo of a young woman laughing in a busy ramen shop at night, steam rising from her bowl, neon signs reflected in the window behind her, natural skin texture, slight motion blur, authentic UGC feel

Requested: nano_banana_pro
Backend: nano_banana_2
Job: 44061d98-0271-4f9a-8284-149b9c340555

{"platform":"Higgsfield","aspect_ratio":"16:9","resolution":"2k","quality":"model default","other":"model defaults"}

Requested Nano Banana Pro; Higgsfield reported nano_banana_2. This output is not a confirmed Pro benchmark.

Download full-resolution original · Model context

Create with Nano Banana Pro

Affiliate link: we may earn a commission if you subscribe. Your model choice comes first. How links work

Soul 2.0 — ugc-realism-ramen
ugc realism ramen

Soul 2.0

Editorial observations: The woman laughs over a cropped bowl with visible steam. Skin texture and grain give a phone-photo feel. A red sign and ceiling lights reflect in the window; there is little obvious motion blur.

2048×1152 native · 0.12 credits (preflight) · 2026-10-07

Prompt and generation provenance

Candid smartphone photo of a young woman laughing in a busy ramen shop at night, steam rising from her bowl, neon signs reflected in the window behind her, natural skin texture, slight motion blur, authentic UGC feel

Requested: soul_2
Backend: text2image_soul_v2
Job: 49ab399c-0db5-4e06-afe0-4a4857775e27

{"platform":"Higgsfield","aspect_ratio":"16:9","resolution":"2k","quality":"model default","other":"model defaults"}

Higgsfield-reported backend ID recorded; provider internals not independently verified.

Download full-resolution original · Model context

Create with Soul 2.0

Affiliate link: we may earn a commission if you subscribe. Your model choice comes first. How links work

Continue the decision.

Common questions / direct answers

Before you choose.

Have you tested this exact prompt?

No. This is clearly labeled prompting guidance, grounded in the cited input controls. Generated benchmark outputs will be shown only with complete settings and provenance.

What is the difference between the still persona and the motion stage?

Treat them as separate deliverables. First curate the identity sheet in an image workflow; then use the selected frame/reference set in a video endpoint whose documented roles support it. The same platform does not imply interchangeable model inputs.

Is this a proven way to gain followers or earn money?

No. This page is an editorial production workflow. We have not measured growth, revenue, engagement or consistent identity retention on these prompts.

Sources and verification.

Start with one clear brief.

Verify the current model, variant and credit estimate before generating.

Create with Soul 2.0

Affiliate link: we may earn a commission if you subscribe. Your model choice comes first. How links work