On this page
What is an AI influencer ad?
An AI influencer ad is a short social video in which a fictional, AI-generated person presents, uses or reacts to a product, and the same character appears across many videos like a real creator would. The face, the personality and the look are designed once and then reused.
The difference from a one-off AI UGC clip is reuse. An AI influencer is a recurring character: the same face in the review, the dance trend, the street interview and the get-ready video. The closest public examples are on AI get-ready UGC ads and AI street interview ads; those come from posts on X and were not made on or for Anglr.
Nearly every high-volume search for AI influencers is about tools. This guide covers the method instead, so it works whichever tool you use.
Tools creators use for it
The table counts how many of the reviewed tutorials named each tool. A tutorial that used two tools counts once for each.
Based on 12 public tutorials posted between 4 Jun 2026 and 8 Oct 2026, reviewed by Anglr.
| Tool | What it does in this workflow | Tutorials that used it |
|---|---|---|
| Seedance | Animates the character, from a storyboard or by replacing the person in a reference clip | 4 |
| Higgsfield | Saves the face as a Soul ID; its AI Influencer tool transfers motion onto a locked character | 4 |
| Nano Banana | Makes the realistic portrait and the contact sheet | 2 |
| GPT Image or ChatGPT | Turns a designed character into a realistic photo, or makes the contact sheet | 2 |
Single tutorials also used Midjourney to design the character, Arcads (AI actors) for ready-made actors with a voice, and Max Fusion inside Claude for the images. Three of the twelve posts (from @girlinAI, @Kinza_507 and @MushrafAli3593) describe the same production order in near-identical words on consecutive days and appear to come from one platform's promotion, so we treat them as one method, not three independent ones.
How do you make an AI influencer step by step?
You make it in five steps: design the character, make a realistic portrait, lock the identity, choose how the character will move, then put the product in. The order is our synthesis of the tutorials.
1. Design one character with one recognisable trait
Decide who the character is before you generate anything: age range, style, where they live, how they talk, and one visual detail people remember. @rewind02 puts it plainly: "the trick is the character, not the model". His advice is one instantly recognisable feature, such as a distinctive haircut or an unusual moustache.
@David_TornAI defines face, voice and personality together in a single prompt. John Savage designs the character in Midjourney first. Fraser Cottrell starts by imagining the character, then builds the sheet.
For ads, the personality should fit the products the character will sell. A calm, tidy character suits skincare and home goods; a loud, chaotic one suits snacks and apps. Write it down as a short character brief and reuse it.
2. Make a realistic, phone-style portrait
Turn the design into one realistic image that looks shot on a phone. @wizzardouthere asks Nano Banana Pro for natural light, a vertical frame and an iPhone look, and rules out a studio or ring-light finish. John Savage takes his Midjourney design into GPT Image 2 to make it photoreal. @hamzaoudghiri2 lists age, skin texture, outfit and setting in the prompt and asks for a front-camera, unretouched look.
All three aim at the same thing: skin with texture, everyday clothes and light from a window or a lamp. Polished, symmetrical faces read as AI at a glance.
3. Lock the identity
Fix the face so every later video uses the same person. This is the step most tutorials stress.
There are two ways to do it. Fraser Cottrell and @imFarhanAi build a character sheet (Fraser calls it a contact sheet): the same face from several angles, attached to every generation. @hamzaoudghiri2 saves the character as a Soul ID in Higgsfield instead. Deykhan Ten describes Higgsfield's AI Influencer tool as locking the character first and running every video from it.
Higgsfield's help center lists the same options in its own words: Soul ID for a person, a saved Element for a video character, or the character's images attached as references. Whichever you use, never re-describe the face from scratch in a new prompt.
4. Choose how the character moves
The tutorials split into two routes here. Motion transfer is described by three independent authors; the production route comes from one shared method.
Motion transfer. Film a short reference clip of the action (a dance, a reaction, a walk to camera), then have the model replace the person in it with your character. Fraser Cottrell gives Seedance 2.5 the contact sheet and the reference video. John Savage uses a single-angle clip and Higgsfield's motion transfer. Deykhan Ten picks a trending motion or uploads a clip in Higgsfield's tool. This is the fastest route to trend formats.
Use clips you filmed yourself or have the rights to. Swapping a character into someone else's video copies their performance, and a clip of a recognisable person risks a likeness problem.
Production route. @girlinAI, @Kinza_507 and @MushrafAli3593 (in the near-identical posts noted above) build the ad like a shoot: the influencer, then a hero frame (the key still of the ad), then a storyboard, then generation in Seedance. It is the same board-first order the AI commercials guide describes for brand films, and it gives more control over each shot than motion transfer, so it suits product ads with a script.
@imFarhanAi adds a third option: with the character locked, he runs one continuous 30-second shot in Seedance 2.5 for a vlog-style piece. That is one author's method.
5. Put the product in and keep a library
Once the character works, make the product part of the frame: in hand, on the counter, in use. Attach a clean product photo to each generation so the label stays readable. Deykhan Ten points out that one locked character can be tested across formats and angles before a brand books a real creator.
Keep a folder per character with the brief, the sheet or Soul ID, the voice and the outfits. That library is what makes the character an influencer rather than a one-off face. When you write scripts for it, the AI UGC ads guide covers the selfie clip workflow in detail.
Prompt templates for an AI influencer
These templates are ours. The portrait and sheet follow Google's Nano Banana guide: say what the image is for, be specific, and describe what you want rather than what you don't. The video prompt follows Higgsfield's help center: a short shot list with blocks for style, camera, light, action and audio, and speech written as a label, a colon and the line in quotation marks. Build your own in the AI ad prompt builder, or see Higgsfield prompts and Nano Banana prompts.
Portrait (image model):
A vertical phone photo of a fictional social media creator, for a recurring ad character. [Age range] [gender], [the one memorable trait], [hair], [everyday outfit]. Taken on a phone front camera at arm's length in [their room or setting], soft window light, natural skin texture with pores and small marks, relaxed expression. Unretouched, everyday look. Vertical 9:16.
Character sheet (image model, with the portrait attached):
Image 1 is the character. Create a character sheet of exactly this person: front, three-quarter left, three-quarter right and profile, plus one full-body view, on a plain light grey background in even light. Keep the face, hair, skin and the [memorable trait] identical in every view. Nothing else in the frame.
Product clip (Higgsfield shot list, with the Soul ID or sheet attached):
Style and mood: casual phone video, warm and unpolished.
Camera: handheld front camera, small natural drift.
Lighting: daylight from a window on the left.
Action: the character from the reference picks up [product], turns the label to the camera, then uses it once.
Character: "[hook line in their voice]"
Character: "[one benefit, said casually]"
Audio: room tone, no music.
Common problems and fixes
| Problem | Fix | Source |
|---|---|---|
| The face changes from video to video | Lock the character once and drive every video from it | Deykhan Ten |
| Face, outfit or room drift over a long shot | Lock the character sheet before generating anything | @imFarhanAi |
| The character looks like a studio model | Ask for natural light, a phone camera and no ring light | @wizzardouthere |
| The character is forgettable | Give them one trait people recognise instantly | @rewind02 |
| A single prompt gives an ad you can't direct | Build an influencer, hero frame and storyboard first | @girlinAI |
| The model ignores the reference image | Avoid contradictions between the prompt and the input image | Higgsfield help center |
Rules to follow with AI influencers
Keep the character fictional. Do not base it on a real person or a celebrity, and do not swap it into footage of a recognisable person. Label the ads: on TikTok, AI-generated content needs the AIGC label or your own clear disclaimer under its ad policy, and AI ad labels on Meta and TikTok covers both platforms. Keep every claim to what the brand can legally say: a fictional character describing personal results can mislead viewers.
What creators report
These are the authors' own claims, not results we checked.
- John Savage reports that social accounts run around a single invented character now reach millions of views.
- Deykhan Ten reports that Higgsfield's tool keeps the face the same in every video it makes from a locked character.
If you are just starting, building one character well is the right first project, and the UGC creator guide covers the basics. Anglr is for creators whose ads already perform; if that is you, the creators page explains how Anglr pays for them.
Who this is for
- Independent AI creators and editors. You make the ads yourself. You know your work performs, so you'd rather be paid on what it does than on a flat fee.
- AI agencies and studios. You run a team of editors. You write the brief, your team makes the versions, and you're paid on performance because you know your team's ads scale.
Not for you if
- You only want a flat fee.
- You're new to editing and not yet confident your ads will perform.
- You want every shot spelled out for you.
- You make one style and won't try others.
FAQ
How do you make an AI influencer?
Design a fictional character, generate a realistic portrait of them, and lock the face with a character sheet or a saved identity such as Higgsfield's Soul ID. Then make videos from that locked character, either by transferring motion from a reference clip or by animating a hero frame in a video model.
How do you keep an AI influencer's face the same in every video?
Attach the same multi-angle character sheet, or the same saved identity, to every generation, and never re-describe the face from scratch. The tutorials we reviewed agree that a new prompt for each clip is what makes the person drift.
Can you use an AI influencer in ads?
Yes, as long as the character is fictional, the claims are ones the brand can legally make and the ad carries the AI label each platform requires. On TikTok that means the AIGC label or your own clear disclaimer, per TikTok's ad policy.
Can an AI influencer look like a real person?
It should not. Build a character that does not resemble any real person or celebrity, and use reference videos you filmed yourself or have the rights to. Likenesses of real people are excluded from every workflow on this page.
What tools do you need to make an AI influencer?
An image model for the portrait and character sheet (Nano Banana, GPT Image or Midjourney in the tutorials we reviewed) and a video model or platform for motion (Seedance or Higgsfield most often). One tutorial used Arcads, which generates the creator and their voice from a prompt.
Sources
- @MushrafAli3593 (4 Jun 2026): an AI influencer ad built like a production: influencer, hero frame, storyboard, then Seedance
- @Kinza_507 (5 Jun 2026): the same production order, step by step instead of one lucky prompt
- @girlinAI (5 Jun 2026): influencer, hero frame and storyboard before Seedance 2.0, the way an agency works
- @David_TornAI (11 Jul 2026): one prompt that sets a creator's face, voice and personality, then videos in Arcads
- @wizzardouthere (25 Jul 2026): a phone-style model portrait in Nano Banana Pro with natural light and no studio look
- @imFarhanAi (17 Sep 2026): lock the influencer and character sheet, then run one continuous shot in Seedance 2.5
- @rewind02 (4 Oct 2026): give the character one detail people recognise in half a second
- John Savage (6 Oct 2026): a Midjourney character made photoreal in GPT Image 2, then swapped into a reference clip with motion transfer
- Fraser Cottrell (6 Oct 2026): a contact sheet plus a reference video in Seedance 2.5 to replace the person with the character
- Deykhan Ten (6 Oct 2026): Higgsfield's AI Influencer tool, which locks a character and transfers motion from a trend clip
- @hamzaoudghiri2 (7 Oct 2026): a realistic Soul character stored as a Soul ID to keep one face across videos
- @hamzaoudghiri2 (8 Oct 2026): a detailed character prompt with age, skin, outfit, setting and a front-camera look
- Higgsfield Help Center โ How to Write a Good AI Prompt
- Google Cloud โ The ultimate Nano Banana prompting guide
- TikTok โ Advertising policies: Misleading and false content
Checked: 12 Oct 2026